diff --git a/.gitignore b/.gitignore
index 71ffb087a5d01c888ba31c55ec8ee8adb8e3a82f..eba503751ecd72c5482cd06f6b81aedd10949b8c 100644
--- a/.gitignore
+++ b/.gitignore
@@ -2,10 +2,11 @@
.DS_Store
# Use wildcards as well
+# *.prn
+
*~
*.mat
*.pyc
-# *.prn
*.aux
*.log
*.dvi
@@ -19,6 +20,8 @@
*.pdf
*.npy
*.o
+*.dll
+*.zip
# Can also ignore all directories and files in a directory.
# tmp/**/*
diff --git a/CFiles/Copy of kalman.c b/CFiles/Copy of kalman.c
deleted file mode 100755
index b53d44e67f6162fcc07843849c030ab5e2c0751f..0000000000000000000000000000000000000000
--- a/CFiles/Copy of kalman.c
+++ /dev/null
@@ -1,2751 +0,0 @@
-/*===============================================================================================================
- * Check $$$ for important notes.
- * Check <<>> for updating DW's new switch code or questions for DW.
- *
- * kalcvf_urw(): the Kalman filter forward prediction specialized for only a univariate random walk (urw) process.
- *
- * State space model is defined as follows:
- * z(t+1) = z(t)+eta(t) (state or transition equation)
- * y(t) = x(t)'*z(t)+eps(t) (observation or measurement equation)
- * where for this function, eta and eps must be uncorrelated; y(t) must be 1-by-1. Note that
- * x(t): k-by-1;
- * z(t): k-by-1;
- * eps(t): 1-by-1 and ~ N(0, sigma^2);
- * eta(t): ~ N(0, V) where V is a k-by-k covariance matrix.
- *
- *
- * Written by Tao Zha, May 2004.
- * Revised, May 2008;
-=================================================================================================================*/
-
-/**
-//=== For debugging purpose.
-if (1)
-{
- double t_loglht;
-
- t_loglht = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, " %10.5f\n", t_loglht);
-
- fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
- fprintf(FPTR_DEBUG, "\n wP0_dv:\n");
- WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "\n Vt_dc->C[sti_v=%d]:\n", sti_v);
- WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
- fflush(FPTR_DEBUG);
-}
-/**/
-
-
-#include "kalman.h"
-
-
-static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
-static int Updatekalfilms_1stapp(int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv) //, int storeZ, int storeV)
-{
- int _i;
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
- //---
- TSkalcvfurw *kalcvfurw_ps = tzMalloc(1, TSkalcvfurw);
-
-
- kalcvfurw_ps->indx_tvsigmasq = tv;
- kalcvfurw_ps->fss = T;
- kalcvfurw_ps->kx = k;
-
- //===
- kalcvfurw_ps->V_dm = CreateMatrix_lf(k, k);
- kalcvfurw_ps->ylhtran_dv = CreateVector_lf(T);
- kalcvfurw_ps->Xrhtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->z10_dv = CreateVector_lf(k);
- kalcvfurw_ps->P10_dm = CreateMatrix_lf(k, k);
-
- kalcvfurw_ps->zupdate_dv = CreateVector_lf(k);
- kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->ylhtranpred_dv = CreateVector_lf(T);
- //
- rows_iv = CreateVector_int(T);
- cols_iv = CreateVector_int(T);
- for (_i=T-1; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // if (!storeZ) kalcvfurw_ps->Zpredtran_dm = (TSdmatrix *)NULL;
- // else kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- // if (!storeV) kalcvfurw_ps->Ppred_dc = (TSdcell *)NULL;
- // else {
- // rows_iv = CreateVector_int(T);
- // cols_iv = CreateVector_int(T);
- // for (_i=T; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- // kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // }
-
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
- return (kalcvfurw_ps);
-}
-
-TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps)
-{
- if (kalcvfurw_ps) {
- DestroyMatrix_lf(kalcvfurw_ps->V_dm);
- DestroyVector_lf(kalcvfurw_ps->ylhtran_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Xrhtran_dm);
- DestroyVector_lf(kalcvfurw_ps->z10_dv);
- DestroyMatrix_lf(kalcvfurw_ps->P10_dm);
-
- DestroyVector_lf(kalcvfurw_ps->zupdate_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Zpredtran_dm);
- DestroyCell_lf(kalcvfurw_ps->Ppred_dc);
- DestroyVector_lf(kalcvfurw_ps->ylhtranpred_dv);
-
- free(kalcvfurw_ps);
- return ((TSkalcvfurw *)NULL);
- }
- else return (kalcvfurw_ps);
-}
-
-
-void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps)
-{
- //See the notes of SWZ regarding the government's updating of the parameters in their Phillips-curve equation.
- //NOTE: make sure that the value of kalcvfurw_ps->sigmasq and other input values are given.
- int ti;
- double workd, workdenominv;
- //---
- int fss, kx;
- double sigmasq_fix = kalcvfurw_ps->sigmasq;
-// double sigmasq;
- TSdmatrix *V_dm;
- TSdmatrix *Zpredtran_dm;
- TSdcell *Ppred_dc;
- TSdvector *ylhtran_dv;
- TSdmatrix *Xrhtran_dm;
- //===
- TSdvector *workkxby1_dv = NULL; //kx-by-1.
-// TSdvector *work1kxby1_dv = NULL; //kx-by-1.
- TSdmatrix *workkxbykx_dm = NULL; //kx-by-kx symmetric and positive positive.
-// //===
-// TSdvector *zbefore_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vbefore_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
-// TSdvector *zafter_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vafter_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
- //******* WARNING: Some dangerous pointer movement to gain efficiency *******
-// double *yt_p;
-// double *Vbefore_p;
-// double *Vafter_p;
- TSdvector xt_sdv;
- TSdvector zbefore_sdv;
- //TSdmatrix Vbefore_sdm;
- TSdvector zafter_sdv;
- //TSdmatrix Vafter_sdm;
-
-
- if (!kalcvfurw_ps) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps must be created");
- if (!kalcvfurw_ps->V_dm || !kalcvfurw_ps->ylhtran_dv || !kalcvfurw_ps->Xrhtran_dm || !kalcvfurw_ps->z10_dv || !kalcvfurw_ps->P10_dm)
- fn_DisplayError(".../kalcvf_urw(): input arguments kalcvfurw_ps->V_dm, kalcvfurw_ps->ylhtran_dv, kalcvfurw_ps->Xrhtran_dm, kalcvfurw_ps->z10_dv, kalcvfurw_ps->P10_dm must be given legal values");
- if (!(kalcvfurw_ps->P10_dm->flag & (M_SU | M_SL))) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps->P10_dm must be symmetric");
- fss = kalcvfurw_ps->fss;
- kx = kalcvfurw_ps->kx;
- V_dm = kalcvfurw_ps->V_dm;
- Zpredtran_dm = kalcvfurw_ps->Zpredtran_dm;
- Ppred_dc = kalcvfurw_ps->Ppred_dc;
- ylhtran_dv = kalcvfurw_ps->ylhtran_dv;
- Xrhtran_dm = kalcvfurw_ps->Xrhtran_dm;
- //---
- xt_sdv.n = kx;
- xt_sdv.flag = V_DEF;
- zbefore_sdv.n = kx;
- zbefore_sdv.flag = V_DEF;
- zafter_sdv.n = kx;
- zafter_sdv.flag = V_DEF;
-
- //=== Memory allocation.
- workkxby1_dv = CreateVector_lf(kx);
- workkxbykx_dm = CreateMatrix_lf(kx, kx);
-
-
- //------- The first period (ti=0). -------
- zbefore_sdv.v = kalcvfurw_ps->z10_dv->v;
- zafter_sdv.v = Zpredtran_dm->M;
- xt_sdv.v = Xrhtran_dm->M;
- //---
-
- workd = ylhtran_dv->v[0] - (kalcvfurw_ps->ylhtranpred_dv->v[0]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, kalcvfurw_ps->P10_dm, &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
-
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(kalcvfurw_ps->z10_dv->v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t];
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
-
-
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[0], kalcvfurw_ps->P10_dm, workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[0]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
-
- //------- The rest of the periods (ti=1:T-1). -------
- for (ti=1; ti<fss; ti++) {
- //NOTE: ti=0 has been taken care of outside of this loop.
- zbefore_sdv.v = Zpredtran_dm->M + (ti-1)*kx;
- zafter_sdv.v = Zpredtran_dm->M + ti*kx;
- xt_sdv.v = Xrhtran_dm->M + ti*kx;
- //---
- workd = ylhtran_dv->v[ti] - (kalcvfurw_ps->ylhtranpred_dv->v[ti]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, Ppred_dc->C[ti-1], &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(zbefore_sdv.v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[ti], Ppred_dc->C[ti-1], workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[ti]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
- }
- CopyVector0(kalcvfurw_ps->zupdate_dv, &zafter_sdv);
- Zpredtran_dm->flag = M_GE;
- kalcvfurw_ps->ylhtranpred_dv->flag = V_DEF;
-
-// DestroyVector_lf(zbefore_dv);
-// DestroyMatrix_lf(Vbefore_dm);
-// DestroyVector_lf(zafter_dv);
-// DestroyMatrix_lf(Vafter_dm);
-
- DestroyVector_lf(workkxby1_dv);
-// DestroyVector_lf(work1kxby1_dv);
- DestroyMatrix_lf(workkxbykx_dm);
-}
-
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- General constant (known-time-varying) Kalman filter for DSGE models.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T)
-{
- int _i;
- //===
- TSivector *rows_iv = CreateVector_int(T);
- TSivector *cols_iv = CreateVector_int(T);
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfiltv_tag *kalfiltv_ps = tzMalloc(1, struct TSkalfiltv_tag);
-
-
- //--- Default value.
- kalfiltv_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfiltv_ps->ny = ny;
- kalfiltv_ps->nz = nz;
- kalfiltv_ps->T = T;
-
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfiltv_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfiltv_ps->at_dm = CreateMatrix_lf(ny, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = nz;
- }
- rows_iv->flag = cols_iv->flag = V_DEF;
- kalfiltv_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->bt_dm = CreateMatrix_lf(nz, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- kalfiltv_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->z0_dv = CreateVector_lf(nz);
- kalfiltv_ps->P0_dm = CreateMatrix_lf(nz, nz);
-
-
- //---
- kalfiltv_ps->zt_tm1_dm = CreateMatrix_lf(nz, T);
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Pt_tm1_dc = CreateCell_lf(rows_iv, cols_iv);
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfiltv_ps);
-
-}
-//---
-struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- if (kalfiltv_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfiltv_ps->yt_dm);
- DestroyMatrix_lf(kalfiltv_ps->at_dm);
- DestroyCell_lf(kalfiltv_ps->Ht_dc);
- DestroyCell_lf(kalfiltv_ps->Rt_dc);
- DestroyCell_lf(kalfiltv_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfiltv_ps->bt_dm);
- DestroyCell_lf(kalfiltv_ps->Ft_dc);
- DestroyCell_lf(kalfiltv_ps->Vt_dc);
- //---
- DestroyVector_lf(kalfiltv_ps->z0_dv);
- DestroyMatrix_lf(kalfiltv_ps->P0_dm);
- //---
- DestroyMatrix_lf(kalfiltv_ps->zt_tm1_dm);
- DestroyCell_lf(kalfiltv_ps->Pt_tm1_dc);
-
-
- //---
- tzDestroy(kalfiltv_ps); //Must be freed last!
-
- return ((struct TSkalfiltv_tag *)NULL);
- }
- else return (kalfiltv_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //=== Default value.
- kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
- kalfilmsinputs_1stapp_ps->ztm1_track = -1;
- kalfilmsinputs_1stapp_ps->dtm1_track = -1;
-
- //--- Other key assignments.
- kalfilmsinputs_1stapp_ps->ny = ny;
- kalfilmsinputs_1stapp_ps->nz = nz;
- kalfilmsinputs_1stapp_ps->nst = nst;
- kalfilmsinputs_1stapp_ps->T = T;
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- //--- For output arguments.
- rows_iv = CreateConstantVector_int(T+1, nz);
- cols_iv = CreateConstantVector_int(T+1, nst);
- kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_1stapp_ps);
-}
-//---
-struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- if (kalfilmsinputs_1stapp_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
- //---
- tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
- }
- else return (kalfilmsinputs_1stapp_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_tag *kalfilmsinputs_ps = tzMalloc(1, struct TSkalfilmsinputs_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //--- Default value.
- kalfilmsinputs_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfilmsinputs_ps->ny = ny;
- kalfilmsinputs_ps->nz = nz;
- kalfilmsinputs_ps->nRc = nRc;
- kalfilmsinputs_ps->nRstc = nRstc;
- kalfilmsinputs_ps->nRv = nRv;
- kalfilmsinputs_ps->indxIndRegimes = indxIndRegimes;
- kalfilmsinputs_ps->T = T;
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_ps->at_dm = CreateMatrix_lf(ny, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, ny);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, ny);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_ps->bt_dm = CreateMatrix_lf(nz, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, nz);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- if (indxIndRegimes)
- {
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRc*nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- //--- For output arguments.
- if (indxIndRegimes)
- {
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRc*nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_ps);
-
-}
-//---
-struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps)
-{
- if (kalfilmsinputs_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->z0_dm);
- DestroyCell_lf(kalfilmsinputs_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_ps->Pt_tm1_d4);
- //---
- tzDestroy(kalfilmsinputs_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_tag *)NULL);
- }
- else return (kalfilmsinputs_ps);
-};
-
-
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//-- Constant-parameters (known-time-varying) Kalman filter
-//-----------------------------------------------------
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //General constant (known-time-varying) Kalman filter for DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- int T = kalfiltv_ps->T;
- int Tp1 = T + 1;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- int indx_badlh = 0; //1: bad likelihood with, say, -infinity of the LH value.
- int tdata, ti;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv; //double loglh_tdata; //logdetDtdata.
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- //--- Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
-
-
-
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "Fatal error: tz_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[0], P0_dm);
-
- //====== See p.002 in LiuWZ. ======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++ )
- {
- //Base-0 timing.
- ti = tdata + 1; //Next period.
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (kalfiltv_ps->loglh = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh += -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //loglh += -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (ti<T)
- {
- //Updating only up to tdata=T-2. The values at ti=T or tdata=T-1 will not be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[ti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + ti*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[ti], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + ti*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[ti], Vt_dc->C[ti]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[ti], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[ti], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[ti], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[ti], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (kalfiltv_ps->loglh = loglh);
-}
-/**
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //This function is used to test tz_logTimetCondLH_kalfiltv().
- int T = kalfiltv_ps->T;
- int tdata;
- double loglh;
-
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++) loglh += tz_logTimetCondLH_kalfiltv(0, tdata+1, kalfiltv_ps);
-
- return (loglh);
-}
-/**/
-//-----------------------------------------------------
-//-- Updating Kalman filter at time t for constant-parameters (or known-time-varying) Kalman filter.
-//-----------------------------------------------------
-double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //st: base-0 grand regime at time t, which is just a dummy for this constant-parameter function in order to use
- // Waggoner's automatic functions.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
- //
- //log LH at time t for constant (known-time-varying) Kalman-filter DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood at time t.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // April 2008, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //--- Output arguments.
- double loglh_timet; //log likelihood at time t.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
- //--- Input arguments.
- int tdata, tp1;
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- int T = kalfiltv_ps->T;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
-
-
- tdata = (tp1=inpt) - 1; //Base-0 time.
-
- //======= Initial condition. =======
- if (tdata==0)
- {
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_logTimetCondLH_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(FPTR_DEBUG, "Fatal error: tz_logTimetCondLH_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and thus the initial conditions must be supplied by, say, input arguments");
- fflush(FPTR_DEBUG);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[tdata], P0_dm);
- }
-
-
- //======= Liklihood at time t (see p.002 in LiuWZ). =======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //======= Updating for the next period. =======
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (tp1<T)
- {
- //Updating only up to tdata=T-2, because the values at tp1=T or tdata=T-1 will NOT be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[tp1], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + tp1*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[tp1], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + tp1*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[tp1], Vt_dc->C[tp1]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[tp1], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[tp1], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[tp1], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[tp1], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (loglh_timet);
-}
-
-
-
-
-//-----------------------------------------------------
-//- WARNING: bedore using this function, make sure to call the following functions
-// Only once in creating lwzmodel_ps: Refresh_kalfilms_*(lwzmodel_ps);
-// Everytime when parameters are changed: RefreshEverything(); RefreRunningGensys_allcases(lwzmodel_ps) in particular.
-//-----------------------------------------------------
-double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //st: base-0 grand regime -- deals with the cross-section values at time t.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
- //-- Output arguments
- double loglh_timet;
- //--- Input arguments
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- Local variables
- int tbase0;
- double logdet_Dtdata;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- TSdvector etdata_sdv;
- //=== Work arguments.
- TSdvector *wny_dv = CreateVector_lf(ny);
-
-
-
- //--- Critical checking.
- if (inpt > kalfilmsinputs_1stapp_ps->T)
- fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): The time exceeds the\n"
- " data sample size allocated the structure TSkalfilmsinputs_1stapp_tag");
-
- //--- The following is for safe guard. InitializeKalman_z10_P10() should be called in, say, RefreshEverything().
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- if (!InitializeKalman_z10_P10(kalfilmsinputs_1stapp_ps, (TSdmatrix *)NULL, (TSdcell *)NULL))
- fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): the system is non-stationary when calling"
- " InitializeKalman_z10_P10(). Please call this function in RefreshEverthing() and"
- " set the likehood to be -infty for early exit");
-
- tbase0=inpt-1;
-
- //------------------- The order matters. Updatekalfilms_1stapp() must be called before Update_et_Dt_1stapp(). -----------------
- //--- $$$ Critical updating where we MUSt have inpt-1. If inpt, Updatekalfilms_1stapp() will call this function again
- //--- $$$ because DW function ProbabilityStateConditionalCurrent() need to access this function at time inpt,
- //--- $$$ which has not computed before Updatekalfilms_1stapp(). Thus, we'll have an infinite loop.
- Updatekalfilms_1stapp(tbase0, kalfilmsinputs_1stapp_ps, smodel_ps);
-// //--- $$$ Critical updating.
-// Update_et_Dt_1stapp(tbase0, kalfilmsinputs_1stapp_ps);
-// //This function will give Dtdata_d4->F[tbase0], etdata_dc->C[tbase0], and PHtran_tdata_d4->F[tbase0].
-
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- //--- Forming the log conditional likelihood at t.
- etdata_sdv.n = ny;
- etdata_sdv.v = etdata_dc->C[tbase0]->M + ny*st;
- etdata_sdv.flag = V_DEF;
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_d4->F[tbase0]->C[st]))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, &etdata_sdv, '/', Dtdata_d4->F[tbase0]->C[st]);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, &etdata_sdv);
- //Done with all w*_dv.
-
- //===
- DestroyVector_lf(wny_dv);
-
- return (loglh_timet);
-}
-//======================================================
-//= Computing z_{1|0} and P_{1|0} for each new parameter values.
-//======================================================
-int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc)
-{
- //See p.001 and p.004 in LWZ Model II.
- //Outputs:
- // return 1: success in initializing; 0: initializing fails, so the likelihood must be set to -infty outside this function.
- // ztm1_track to track the time up to which Kalman filter have been updated.
- // z0_dm, zt_tm1_dc->C[0]
- // P0_dc, Pt_tm1_d4->F[0]
-
- //--- Output arguments
- TSdmatrix *z0_0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdmatrix *z0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdcell *P0_dc = kalfilmsinputs_1stapp_ps->P0_dc; //nz-by-nz-by-nst.
- //+ Used to get zt_tm1_dc->C[0] and Pt_tm1_d4->F[0] only.
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Input arguments
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //--- Local variables
- int sti;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv, z0_0_sdv, bt_sdv;
- TSdvector yt_sdv, at_sdv;
- //--- For the initial conditions: eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- //===
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
-
-
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- {
- z0_sdv.n = z0_0_sdv.n = bt_sdv.n = nz;
- z0_sdv.flag = z0_0_sdv.flag = bt_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (!kalfilmsinputs_1stapp_ps->indxIni)
- {
- z0_0_dm->flag = z0_dm->flag = M_GE;
- for (sti=nst-1; sti>=0; sti--)
- {
- if (kalfilmsinputs_1stapp_ps->DiffuseScale) //Diffuse initial conditions are used.
- {
- //--- Diffuse condition for z0_dv.
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
- bt_sdv.v = bt_dm->M + bt_sdv.n*sti;
- InitializeConstantVector_lf(&z0_0_sdv, 0.0);
- MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
- VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
- //--- Diffuse condition for P0_dm.
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, kalfilmsinputs_1stapp_ps->DiffuseScale); //To be used for DiffuseScale*I(nz)
- CopyMatrix0(P0_dc->C[sti], Wnzbynz_dm);
- //Done with W*_dm.
- }
- else //Unconditional moments for initial conditions are used.
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti]);
- if (errflag) fn_DisplayError("kalman.c/InitializeKalman_z10_P10(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0-SQRTEPSILON)) //(1.0+EPSILON))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,sti))\b(:,sti);
- z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti]);
- CopySubmatrix2vector(&z0_0_sdv, 0, bt_dm, 0, sti, bt_dm->nrows);
- bdivA_rgens(&z0_0_sdv, &z0_0_sdv, '\\', Wnzbynz_dm);
- //- Under the assumption s_0 = s_1 (this is a short-cut).
- MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
- VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,sti),F(:,:,sti)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti], Ft_dc->C[sti]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- if (0) //0: no printing.
- {
- #if defined (USE_DEBUG_FILE)
- fprintf(FPTR_DEBUG, "\n-------WARNING: ----------\n");
- fprintf(FPTR_DEBUG, "\nIn grand regime sti=%d\n", sti);
- fprintf(FPTR_DEBUG, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #else
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn grand regime sti=%d\n", sti);
- fprintf(stdout, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #endif
- }
- //=== See p.000.3 in LWZ Model II.
- //=== Do NOT use the following option. It turns out that this will often generate explosive conditional liklihood
- //=== at the end of the sample, because Pt_tm1 shrinks to zero overtime due to the sigularity of
- //=== the initila condition P_{1|0}.
- //--- Letting z0_dv = 0.0
- // z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- // InitializeConstantVector_lf(&z0_sdv, 0.0);
- // //--- Letting P0_dm = V
- // CopyMatrix0(P0_dc->C[sti], Vt_dc->C[sti]);
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (0); //Early exit with kalfilmsinputs_1stapp_ps->ztm1_track continues to be -1.
- }
- }
- }
- }
- else
- {
- if (!z10_dm) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition z_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyMatrix0(z0_dm, z10_dm);
-
- if (!P10_dc) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition P_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyCell0(P0_dc, P10_dc);
- }
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t-1 = 1.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t-1 = 1.
-
-
- kalfilmsinputs_1stapp_ps->ztm1_track = 0; //Must reset to 0, meaning initial setting is done and ready for computing LH at t = 1.
-
- Update_et_Dt_1stapp(0, kalfilmsinputs_1stapp_ps);
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (1);
- }
- else
- {
- fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): calling this function makes sense only if"
- " kalfilmsinputs_1stapp_ps->ztm1_track is -1. Please check this value.");
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (0);
- }
-}
-//======================================================
-//= Integrating out the lagged regimes in order to
-//= updating zt_tm1 and Pt_tm1 for next perid tp1 through Kim-Nelson filter.
-//= tdata representing base-0 t timing, while inpt represents base-1 t timing.
-//
-//= Purpose: for each inpt, we integrate out grand regimes st
-//= only ONCE to prevent the dimension of updated zt_tm1 and Pt_tm1 through Kim-Nelson filter.
-//======================================================
-static int Updatekalfilms_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //Output:
- // tm1update
- // z_{t_1+1|t_1}
- // P_{t_1+1|t_1}
- //Input:
- // t-1: base-1 t timing. Thus t-1=inpt-1.
-
- //--- Local variables
- int stp1i, sti, t_2, t_2p1;
- double prob_previous_regimes;
- //-- Output arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Input arguments
- TSdcell *Gt_dc = kalfilmsinputs_1stapp_ps->Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //+
- TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- int T = kalfilmsinputs_1stapp_ps->T;
- TSdvector z0_sdv;
- TSdvector btp1_sdv;
- TSdvector etdata_sdv;
- //=== Work arguments.
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- //+
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //=== For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- //--- Critical checking.
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- fn_DisplayError(".../kalman.c/Updatekalfilms_1stapp(): Make sure InitializeKalman_z10_P10() is called in the function RefreshEverthing()");
-
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- btp1_sdv.n = nz;
- btp1_sdv.flag = V_DEF;
- //+
- etdata_sdv.n = ny;
- etdata_sdv.flag = V_DEF;
-
- for (t_2=kalfilmsinputs_1stapp_ps->ztm1_track; t_2<t_1; t_2++)
- {
- //If t_1 <= ztm1_track, no updating.
- //If t_1 > ztm1_track, updating z_{t|t-1} and P_{t|t-1} up to t-1 = t_1.
-
- zt_tm1_dc->C[t_2p1=t_2+1]->flag = M_GE;
- for (stp1i=nst-1; stp1i>=0; stp1i--)
- {
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //=== Updating for next period by integrating out sti..
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i], PHtran_tdata_d4->F[t_2]->C[sti], 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_d4->F[t_2]->C[sti]);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- etdata_sdv.v = etdata_dc->C[t_2]->M + ny*sti;
- z0_sdv.v = zt_tm1_dc->C[t_2]->M + nz*sti; //sti: regime at time t_2.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, &etdata_sdv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i], Pt_tm1_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
-
- //--- Integrating out the state at t_2 using
- //--- P(s_t_2|Y_{t_2}, theta) = ProbabilityStateConditionalCurrent(sti, t_2, smodel_ps);
- //--- One can also access to P(s_t_2|Y_{t_2}, theta) by using ElementV(smodel_ps->V[t_2],s_{t_2}i),
- //--- but this access will not call my function logTimetCondLH(), thus no updating for
- //--- P(s_t_2|Y_{t_2}, and thus leading to incorrect results.
- prob_previous_regimes = ProbabilityStateConditionalCurrent(sti, t_2, smodel_ps);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period.
- z0_sdv.v = zt_tm1_dc->C[t_2p1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[t_2p1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- //--- $$$ The following is important because it tells ProbabilityStateConditionalCurrent(), which calls
- //--- $$$ logTimetCondLH_kalfilms_1stapp(), which calls recursively this function again, that there is no
- //--- $$$ need to update Kalman filter for the period before kalfilmsinputs_1stapp_ps->ztm1_track.
- kalfilmsinputs_1stapp_ps->ztm1_track = t_2p1; //Means that z_{t_2p1+1|t_2p1} and P_{t_2p1+1|t_2p1} are done.
-
- //--- $$$ This function must be called after all the above computations are done.
- Update_et_Dt_1stapp(t_2p1, kalfilmsinputs_1stapp_ps);
- }
-
-
- //===
- DestroyMatrix_lf(Wnzbynz_dm);
- //
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (kalfilmsinputs_1stapp_ps->ztm1_track);
-}
-//======================================================
-//= Computes etdata and Dtdata for all grand regimes st at tbase0=inpt-1 or dtm1_track
-//= to prevent recomputing this object for different st at given tbase0.
-//======================================================
-static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- //Output:
- // dtm1_track is updated in this function.
- // PHtran_tdata_d4->F[t-1]
- // etdata_dc->C[t-1]
- // Dtdata_d4->F[t-1]
- //Input:
- // t_1=inpt-1: base-0 timing for et and Dt before the likelihood at time inpt is computed.
-
- //--- Local variables
- int sti, tbase0;
- //-- Output arguments
- TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdcell *yt_tm1_dc = kalfilmsinputs_1stapp_ps->yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0).
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- input arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-T.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-T.
- //+
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv;
- TSdvector yt_sdv, at_sdv;
- TSdvector etdata_sdv, yt_tm1_sdv;
- //=== Work arguments.
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- etdata_sdv.n = yt_tm1_sdv.n = ny;
- etdata_sdv.flag = yt_tm1_sdv.flag = V_DEF;
-
- for (tbase0=(kalfilmsinputs_1stapp_ps->dtm1_track+1); tbase0<=t_1; tbase0++)
- {
- //Note tbase0<=t_1, NOT tbase0<t_1.
- //If t_1 < (dtm1_track+1), no updating.
- //If t_1 >= (dtm1_track+1), updating etdata_dc->C[t-1] and Dtdata_d4->F[t-1] up to t-1=t_1.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti], 1.0, 0.0, 'N', 'T');
- CopyMatrix0(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4->F[tbase0]->C[sti], PHtran_tdata_dm);
-
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti*at_dm->nrows; //grand regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- etdata_sdv.v = etdata_dc->C[tbase0]->M + etdata_sdv.n*sti;
- yt_tm1_sdv.v = etdata_dc->C[tbase0]->M + yt_tm1_sdv.n*sti;
- CopyVector0(&yt_tm1_sdv, &at_sdv);
- MatrixTimesVector(&yt_tm1_sdv, Ht_dc->C[sti], &z0_sdv, 1.0, 1.0, 'N'); //a + H*z_{t|t-1}.
- VectorMinusVector(&etdata_sdv, &yt_sdv, &yt_tm1_sdv); //y_t - a - H*z_{t|t-1}.
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
- CopyMatrix0(Dtdata_d4->F[tbase0]->C[sti], Dtdata_dm); //Saved to be used for logTimetCondLH_kalfilms_1stapp().
- }
-
- //--- $$$ This tracker functions the same way as kalfilmsinputs_1stapp_ps->ztm1_track.
- kalfilmsinputs_1stapp_ps->dtm1_track = tbase0;
- }
-
- //===
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (kalfilmsinputs_1stapp_ps->dtm1_track);
-}
-
-
-
-
-
-
-//-----------------------------------------------------
-//------------ OLD Code --------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//
-//- IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
-//- the coefficient regime is the 1st state variable, and
-//- the volatility regime is the 2nd state variable.
-//-----------------------------------------------------
-#if defined (NEWVERSIONofDW_SWITCH)
-double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int comst_c; //composite (s_tc, s_{t-1}c)
- int st_c, stm1_c, st_v;
- int comsti_c; //composite (s_tc, s_{t-1}c)
- int sti, sti_c, stm1i_c, sti_v;
- int comstp1i_c; //composite (s_{t+1}c, s_tc)
- int stp1i, stp1i_c, stp1i_v;
- int tbase0, tp1;
- double logdet_Dtdata, loglh_timet;
- static int record_tbase1_or_inpt_or_tp1 = 0;
- static int passonce;
- double prob_previous_regimes;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRstc = kalfilmsinputs_ps->nRstc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- int **Index = smodel_ps->sv->index; //Regime-switching states.
- //smodel_ps->sv->index is for our new code.
- // For old code (before 9 April 08 and before dsge_switch is created), use smodel_ps->sv->Index;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_logTimetLH_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- tbase0 = (tp1=inpt) - 1;
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (tbase0==0)
- {
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[comsti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[comsti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, comsti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[comsti_c], Ft_dc->C[comsti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
-//=== ???????? For debugging purpose.
-//if ((inpt<2) && (st==0))
-//{
-// fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
-// fprintf(FPTR_DEBUG, "wP0_dv:\n");
-// WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Vt_dc->C[sti_v=%d]:\n", sti_v);
-// WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
-// fflush(FPTR_DEBUG);
-
-//}
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime comsti_c=%d and sti_v=%d and at time=%d\n", comsti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
- }
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- if (indxIndRegimes )
- {
- if (nRc==1) //Volatility.
- {
- comst_c = st_c = 0;
- st_v = st;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_v = Index[st][1]; //volatility state s_tv
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comst_c = st_c = st;
- st_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = st_c = Index[st][0];
- st_v = Index[st][1];
- }
- }
- else //Syncronized regimes
- {
- if (nRc>nRstc)
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = st_c;
- }
- else
- comst_c = st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[comst_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comst_c*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comst_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comst_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //--- Forming the log conditional likelihood at t.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
-//if ((inpt>82) && (inpt<86) )
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %.16e ");
-
-// fflush(FPTR_DEBUG);
-//}
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
-
-
-//=== ???????? For debugging purpose.
-if (inpt==1)
-{
- double wk1, wk2;
-
- wk1 = logdet_Dtdata;
- wk2 = VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, "logdet_Dtdata = %10.5f\n", wk1);
- fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f\n", wk2);
- fprintf(FPTR_DEBUG, "----- etdata_dv: \n");
- WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- yt_dv: \n");
- WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- at_dv: \n");
- WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- z0_dv: \n");
- WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- Ht_dc->C[comst_c=%d]:\n", comst_c);
- WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-
- fprintf(FPTR_DEBUG, "\n\n");
-
-}
-//
-fprintf(FPTR_DEBUG, " %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//fprintf(FPTR_DEBUG, "------------------------\n");
-//fprintf(FPTR_DEBUG, "st=%d and inpt=%d\n", st, inpt);
-//fprintf(FPTR_DEBUG, "loglh_timet = %10.5f\n", loglh_timet);
-//fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-//WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-////WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-////fprintf(FPTR_DEBUG, "\n");
-////WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>82) && (inpt<86) )
-//if (inpt<2)
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// tmp_Dtdata = logdeterminant(Dtdata_dm);
-// tmp_expterm = VectorDotVector(wny_dv, etdata_dv);
-// fprintf(FPTR_DEBUG, "logdeterminant(Dtdata_dm) = %10.5f;\n", tmp_Dtdata);
-// fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f;\n", tmp_expterm);
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&at_sdv:\n");
-// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&z0_sdv:\n");
-// WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Ht_dc->C[comst_c=%d]:\n",comst_c);
-// WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Rt_dc->C[st_v=%d]:\n", st_v);
-// WriteMatrix(FPTR_DEBUG, Rt_dc->C[st_v], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4->F[tbase0]->C[st = %d]:\n",st);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[st], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-
-
-
-
-//// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "zt_tm1_dc->C[tbase0]:\n");
-//// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-//// //WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-//// //fprintf(FPTR_DEBUG, "\n");
-//// fprintf(FPTR_DEBUG, "bt_dm = [\n");
-//// WriteMatrix(FPTR_DEBUG, bt_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-
-//// fprintf(FPTR_DEBUG, "et:\n");
-//// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "yt_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-//// fprintf(FPTR_DEBUG, "at_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-
-//// for (ki=0; ki<Ht_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ht_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ht_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Ft_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ft_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ft_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Vt_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Vt_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Vt_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-// fflush(FPTR_DEBUG);
-//}
-
-
- //======================================================
- //= Updating zt_tm1 and Pt_tm1 for next perid tp1.
- //= tdata = tbase0 is base-0 timing.
- //======================================================
- if (inpt > record_tbase1_or_inpt_or_tp1) //This condition always satisfies at the 1st period (which is inpt=1).
- {
- passonce = 0;
- record_tbase1_or_inpt_or_tp1 = inpt;
- }
- if (!passonce)
- {
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comstp1i_c = stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_v = Index[stp1i][1]; //volatility state s_tv
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comstp1i_c = stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comstp1i_c = stp1i_c = Index[stp1i][0];
- stp1i_v = Index[stp1i][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = stp1i_c;
- }
- else
- comstp1i_c = stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[comsti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comsti_c*at_dm->nrows; //comsti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comsti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comsti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i_v]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_{t-1}, theta) = ElementV(smodel_ps->Z[inpt],s_{inpt}_i).
- //--- Note tbase0 = inpt-1 because the data in DW code (ElementV) is base-1.
- //--- Note at this point, we cannot access to P(s_t|Y_t, theta) = ElementV(smodel_ps->V[inpt],s_{inpt}_i)
- //--- through DW's code. But we can modify my own code to do this later.
- prob_previous_regimes = ElementV(smodel_ps->Z[inpt],sti);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
- }
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>60) && (inpt<65) ) //if (inpt<5)
-//{
-// int kip1; //Must be declared at the top of this "if" block.
-
-// fprintf(FPTR_DEBUG, "zt_tm1t=[\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-
-// for (ki=0; ki<Pt_tm1_d4->F[tbase0]->ncells; ki++)
-// {
-// kip1 = ki+1;
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4t(:,:,%d)=[\n", kip1);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[ki], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-// }
-
-// fflush(FPTR_DEBUG);
-//}
-
-
-//=== ???????? For debugging purpose.
-fprintf(FPTR_DEBUG, " loglh_timet = %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-#endif
-
-
-/**
-//----------------------------------------------------------------
-//-- Tested OK, but has not use because tz_Refresh_z_T7P_T_in_kalfilms_1st_approx()
-//-- cannot access to ElementV(smodel_ps->V[tp1],sti) or ElementV(smodel_ps->V[tbase0],sti)
-//-- because no likelihood has been formed at all before this function is called.
-//----------------------------------------------------------------
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//-----------------------------------------------------
-void tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- double debug1;
- //--- Local variables
- int stp1i, stp1i_c, stp1i_v, sti, sti_c, sti_v, tbase0, tp1;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, sti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti_c], Ft_dc->C[sti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime sti_c=%d and sti_v=%d and at time=%d\n", sti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
-
-
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[0]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[0]->C[0], " %.16e ");
-
-
- //============== Updating zt_tm1 and Pt_tm1. ==================
- for (tbase0=0; tbase0<T; tbase0++ )
- {
- //tdata = tbase0 is base-0 timing.
- tp1 = tbase0 + 1; //Next period.
-
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if (indxIndRegimes)
- {
- stp1i_c = smodel_ps->sv->Index[stp1i][0];
- stp1i_v = smodel_ps->sv->Index[stp1i][1];
- }
- else
- {
- stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti_c*at_dm->nrows; //sti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[sti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_t, theta) = ElementV(smodel_ps->V[t+1],s_{t+1}_i).
- //--- Note because the data in DW code (ElementV) is base-1, t+1 is actually tbase0.
- debug1 = ElementV(smodel_ps->V[tp1],sti); //?????? Debug.
- //ScalarTimesVectorUpdate(ztp1_dv, ElementV(smodel_ps->V[tp1],sti), ztp1_t_dv);
- //ScalarTimesMatrix(Ptp1_dm, ElementV(smodel_ps->V[tp1],sti), Ptp1_t_dm, 1.0);
- ScalarTimesVectorUpdate(ztp1_dv, 0.5, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, 0.5, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
-
-// fprintf(FPTR_DEBUG, "\n &yt_sdv:\n");
-// WriteMatrix(FPTR_DEBUG, &yt_sdv, " %.16e ");
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[tp1]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tp1], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[tp1]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tp1]->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fflush(FPTR_DEBUG);
-
-
- }
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-}
-//-----------------------------------------------------
-//- Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// (1) RunningGensys_const7varionly(lwzmodel_ps);
-// (2) Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-// (3) tz_Refresh_z_T7P_T_in_kalfilms_1st_approx();
-//- before using kalfilms_timet_1st_approx().
-//-----------------------------------------------------
-double tz_kalfilms_timet_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0. The same for (T+1)-by-1 gbeta_dv and nlcoefs-by-(T+1) galpha_dm.
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int st_c, st_v, tbase0;
- double loglh_timet;
- //--- Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- TSdvector yt_sdv, at_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- if (smodel_ps->sv->nstates != zt_tm1_dc->C[0]->ncols) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Make sure that the column dimension of zt_tm1_dc->C is the same as smodel_ps->sv->nstates");
-
- tbase0 = inpt - 1; //base-0 time t.
-
- if (indxIndRegimes && (nRc==1))
- {
- st_c = 0;
- st_v = st;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- st_c = st;
- st_v = 0;
- }
- else if (indxIndRegimes)
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
- st_c = smodel_ps->sv->Index[st][0];
- st_v = smodel_ps->sv->Index[st][1];
- }
- else
- {
- st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + st_c*at_dm->nrows; //st_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[st_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
-
- //--- Forming the log conditional likelihood at t.
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //===
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-/**/
-
-
-
-
diff --git a/CFiles/Copy of kalman.h b/CFiles/Copy of kalman.h
deleted file mode 100755
index 7e869dd7d9575d1f274640b3615a7d77c82ec6cd..0000000000000000000000000000000000000000
--- a/CFiles/Copy of kalman.h
+++ /dev/null
@@ -1,300 +0,0 @@
-#ifndef __KALMAN_H__
- #define __KALMAN_H__
-
- #include "tzmatlab.h"
- #include "mathlib.h"
- #include "switch.h"
- #include "fn_filesetup.h" //Used to call WriteMatrix(FPTR_DEBUG,....).
-
-
- typedef struct TSkalcvfurw_tag {
- //urw: univariate random walk kalman filter. Desigend specially for the 2006 AER SWZ paper.
-
- //=== Input arguments.
- int indx_tvsigmasq; //0: constant siqmasq in Kalman updating (default);
- //1: Keyensian (project-specific) type of time-varying sigmasq in Kalman updating; See pp.37 and 37a in SWZ Learning NOTES;
- //2: project-specific type;
- //3: another project-specific type.
- double sigmasq; //Variance for the residual eps(t) of the measurement equation.
- int fss; //T: effective sample size (excluding lags).
- int kx; //dimension for x(t).
- TSdmatrix *V_dm; //kx-by-kx. Covariance (symmetric and positive definite) matrix for the residual eta(t) of the transition equation.
- TSdvector *ylhtran_dv; //1-by-T of y(t). The term lh means lelf hand side and tran means transpose.
- TSdmatrix *Xrhtran_dm; //kx-by-T of x(t). The term rh means right hand side and tran means transpose.
- TSdvector *z10_dv; //kx-by-1. Initial condition for prediction: z_{1|0}.
- TSdmatrix *P10_dm; //kx-by-kx symmetric matrix. Initial condition for the variance of the prediction: P_{1|0}.
-
- //=== Output arguments.
- TSdvector *zupdate_dv; //kx-by-1. z_{T+1|T}.
- TSdmatrix *Zpredtran_dm; //kx-by-T matrix of one-step predicted values of z(t). [z_{2|1}, ..., z_{t+1|t}, ..., z_{T+1|T}].
- //Set to NULL (no output) if storeZ = 0;
- TSdcell *Ppred_dc; //T cells and kx-by-kx symmetric and positive definite matrix for each cell. Mean square errors of predicted state variables.
- //{P_{2|1}, ..., P{t+1|t}, ..., P{T+1|T}. Set to NULL (no output if storeV = 0).
- TSdvector *ylhtranpred_dv; //1-by-T one-step prediction of y(t) or ylhtran_dv. Added 03/17/05.
-
- //=== Function itself.
- void (*learning_fnc)(struct TSkalcvfurw_tag *, void *);
- } TSkalcvfurw; //urw: univariate random walk.
- //
- typedef void TFlearninguni(struct TSkalcvfurw_tag *, void *); //For linear rational expectations models.
-
-
- //=== Better version is TSkalfilmsinputs_1stapp_tag. Kalman filter for constant or known-time-varying DSGE models.
- typedef struct TSkalfiltv_tag
- {
- //General (known-time-varying) Kalman filter for DSGE models.
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Not used if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as a initial condition
- // and with z_{t+1|t} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as a initial condition
- // and with P_{t+1|t} as the last element. Thus, we can use it as a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-T.
- TSdcell *Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-T.
- TSdcell *Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv; //nz-by-1;
- TSdmatrix *P0_dm; //nz-by-nz.
-
- //=== Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc; //nz-by-nz-T.
- } TSkalfiltv;
-
-
-
- //=== Inputs for filter for Markov-switching DSGE models at any time t.
- typedef struct TSkalfilmsinputs_1stapp_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on the grand regime s_t that follows a Markov-chain process
- // and is taken as given.
- //
- // Inputs at time t are as follows where nst is number of grand regimes (including lagged regime
- // and coefficients and shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nst matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nst 3-D of Markov-switching matrices in the measurement equation.
- // R is an n_y-by-n_y-by-nst 3-D of Markov-switching covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nst 3-D of Markov-switching E(eta_t * eps_t').
- // ------
- // b is an n_z-by-nst matrix of Markov-switching input vectors in the state equation with b(:,st) as an initial condition.
- // (alternatively, with the ergodic weighted b(:,st) as an initial condition).
- // F is an n_z-by-n_z-by-nst 3-D of Markov-switching transition matrices in the state equation with F(:,:,st)
- // as an initial condition (alternatively, with the ergodic weighted F(:,:,st) as an initial condition).
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation
- // with V(:,:,st) as an initial condition (alternatively, with the ergodic weighted V(:,:,st) as an initial condition).
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nst matrix of initial condition (Not used if indxIni=0).
- // P0 is an n_z-by-n_z-by-nst 3-D of initial condition (Not used if indxIni=0).
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,st))\b(:,st)
- // vec(P0_0m1) = (I-kron(F(:,:,st),F(:,:,st)))\vec(V(:,:,st))
- // Note that all eigenvalues of the matrix F(:,:,st) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha. Revised, April 2008.
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nst; //number of grand composite regimes (current and past regimes, coefficient and volatility regimes).
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0,
- //0: using the unconditional momnets for any given regime at time 0 (default when indxDiffuse = 0).
- int indxDiffuse; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- double DiffuseScale; //A large (infinity) number when indxDiffuse = 1.
- int ztm1_track; //t-1 = -1: no initial conditions z_{1|0} and P_{1|0} has been computed yet, but will be using InitializeKalman_z10_P10(),
- //t-1 >= 0:T-1: z_{t|t-1} and P_{t|t-1} are updated up to t-1.
- int dtm1_track; //t-1 = -1: no etdata_dc->C[0] or Dtdata_d4->F[0] has been computed yet.
- //t-1 >= 0:T-1: etdata_dc->C[t-1] and Dtdata_d4->F[t-1] are updated up to t-1.
-
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nst.
- TSdcell *Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nst.
- TSdcell *Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_0_dm; //nz-by-nst. z_{0|0}.
- TSdmatrix *z0_dm; //nz-by-nst. z_{1|0}.
- TSdcell *P0_dc; //nz-by-nz-by-nst. P_{1|0}
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nst-by-(T+1), where z_{1|0} is an initial condition (1st element with t-1=0 or t=1 for base-1) and
- // the terminal condition z_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, z_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1), where P_{1|0} is an initial condition (1st element with t-1=0) and
- // the terminal condition P_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, P_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- //+ Will be save for updating likelihood and Kalman filter Updatekalfilms_1stapp(), so save time to recompute these objects again.
- TSdfourth *PHtran_tdata_d4; //nz-by-ny-by-nst-T, P_{t|t-1}*H_t'. Saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc; //ny-by-nst-by-T (with base-0 T), forecast errors e_t in the likelihood.
- TSdcell *yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0). Incorrect now (Used to back out structural shocks).
- TSdfourth *Dtdata_d4; //ny-by-ny-nst-by-T, forecast covariance D_t in the likelihood. Saved for updating Kalman filter Updatekalfilms_1stapp().
- } TSkalfilmsinputs_1stapp;
-
-
- //=== OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
- typedef struct TSkalfilmsinputs_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs at time t are as follows where nRc is number of regimes for coefficients
- // nRv is number of regimes for volatility (shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nRc matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nRc 3-D of Markov-switching matrices in the measurement equation.
- // R is an n_y-by-n_y-by-nRv 3-D of Markov-switching covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nRv 3-D of Markov-switching E(eta_t * eps_t').
- // ------
- // b is an n_z-by-nRc matrix of Markov-switching input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-nRc 3-D of Markov-switching transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIndRegimes: 1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nRc*nRv matrix of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // z0 is an n_z-by-nRv matrix of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRc*nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nRc; //number of composite regimes (current and past regimes) for coefficients.
- int nRstc; //number of coefficient regimes at time t.
- int nRv; //number of regimes for volatility (shock variances).
- int indxIndRegimes; //1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nRc.
- TSdcell *Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm; //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- TSdcell *P0_dc; //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- } TSkalfilmsinputs;
-
-
-
-
- //--- Functions for univariate random walk kalman filter.
- TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv); //, int storeZ, int storeV);
- TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps);
- void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps);
-
- //--- New Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T);
- struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
- int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc);
- double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-
- //--- OLD Code: Functions for general constant Kalman filter.
- struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T);
- struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- //Used to test tz_logTimetCondLH_kalfiltv(). (Done April 08). double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps);
-
- //--- OLD Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T);
- struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps);
- double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps);
- //IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
- // the coefficient regime is the 1st state variable, and
- // the volatility regime is the 2nd state variable.
-#endif
-
diff --git a/CFiles/congradmin.c b/CFiles/congradmin.c
old mode 100755
new mode 100644
diff --git a/CFiles/congradmin.h b/CFiles/congradmin.h
old mode 100755
new mode 100644
diff --git a/CFiles/congradminOrigWorks.c b/CFiles/congradminOrigWorks.c
deleted file mode 100755
index e4cbb6e7d6e1f95f365b3e72d91b42a6da0c5588..0000000000000000000000000000000000000000
--- a/CFiles/congradminOrigWorks.c
+++ /dev/null
@@ -1,557 +0,0 @@
-/*************************************************************
- * Conjugate Gradient Minimization Methods. See Numerical Recipes in C by Press, Flannery, Teukolsky, and Vetterling.
- * (I) frprmn(): Plolak-Ribiere method with the line minimization without using the derivative information.
- * (II) dlinmin(): Fletcher-Reeves method with the line minimization using the derivative information.
- *
- * Modified by Tao Zha, September 2003.
-*************************************************************/
-
-#include "congradmin.h"
-
-static void linmin(double p[], double xi[], int n, double *fret, double tol_brent, int itmax_brent, double (*func)(double [], int));
-static double brent(double ax, double bx, double cx, double (*f)(double), double tol_brent, double itmax_brent, double *xmin);
-//
-static void dlinmin(double p[], double xi[], int n, double *fret, double tol_dbrent, double itmax_dbrent, double *grdh_p, double (*func)(double [], int), void (*dfunc)(double [], double [], int, double (*func)(double [], int), double *, double));
-static double dbrent(double ax, double bx, double cx, double (*f)(double), double (*df)(double, double *), double *grdh_p, double tol_dbrent, double itmax_dbrent, double *xmin);
-static double df1dim(double x, double *grdh_p);
-//
-static void mnbrak(double *ax, double *bx, double *cx, double *fa, double *fb, double *fc, double (*func)(double));
-static double f1dim(double x);
-//
-static double ftd_norm2(double *vnew_p, double *vold_p, int _n);
-
-
-
-#define STRLEN 192
-static FILE *fptr_interesults = (FILE *)NULL; //Printing intermediate results to a file.
-static char filename_sp3vecs[STRLEN]; //Three vectors. 1st row: line search direction; 2nd row: numerical gradient; 3rd row: vectorized parameters.
-//static FILE *fptr_interesults_db = (FILE *)NULL; //Printing intermediate results to a file for debugging (db).
-#define PRINTON //Added by TZ, September 2003.
-#define EPS 1.0e-10 //Small number to rectify special case of converging to exactly zero function value.
-#ifdef PRINTON //Added by TZ, September 2003.
- #define FREEALL {tzDestroy(xi); tzDestroy(h); tzDestroy(g); tzDestroy(pold); tzDestroy(numgrad)}
-#else
- #define FREEALL {tzDestroy(xi); tzDestroy(h); tzDestroy(g);}
-#endif
-void frprmn(double p[], int n, int *iter, double *fret,
- double (*func)(double [], int), void (*dfunc)(double [], double [], int, double (*func)(double [], int), double *, double),
- double *ftol_p, int *itmax_p, double *tol_brent_p, int *itmax_brent_p, double *grdh_p) {
- //Outputs:
- // p[0, ..., n-1]: the location of the minimum if it converges, which replaces the starting value.
- // iter: pointer to the number of iterations that were performed.
- // fret: pointer to the minimum value of the function.
- //Inputs:
- // p[0, ..., n-1]: a starting point for the minimization.
- // n: the dimension of p.
- // ftol_p: pointer to the convergence tolerance on the objective function value. Default: 1.0e-4 if NULL.
- // itmax_p: pointer to the maximum number of iterations in the main minimization program frprmn(). Default: 2000 if NULL.
- // tol_brent_p: pointer to the convergence tolerance for the line minimization in brent(). Default: 2.0e-4 if NULL.
- // itmax_brent_p: pointer to the maximum number of iterations for the line minimization in brent(). Default: 100 if NULL.
- // grdh: pointer to the user's specified step size for a numerical gradient. If NULL, dfunc() (i.e., gradcd_gen()) will select grdh automatically.
- // func(): the objective function.
- // dfunc(): the gradient function computing the numerical gradient. In the form of gradcd_gen() in cstz.c.
- int j, its, itmax, itmax_brent;
- double gg, gam, fp, dgg, ftol, tol_brent;
- double *g=NULL, *h=NULL, *xi=NULL;
- #ifdef PRINTON //Added by TZ, September 2003.
- time_t begtime, currentime;
- double normforp, *pold = NULL, *numgrad = NULL;
- #endif
-
- //=== Memory allocation.
- g=tzMalloc(n, double);
- h=tzMalloc(n, double);
- xi=tzMalloc(n, double);
- //
- numgrad = tzMalloc(n, double); //Added by TZ, September 2003.
- #ifdef PRINTON //Added by TZ, September 2003.
- pold = tzMalloc(n, double);
- #endif
-
- //=== Default values.
- if (!ftol_p) ftol = 1.0e-4; else ftol = *ftol_p;
- if (!itmax_p) itmax = 200; else itmax = *itmax_p;
- if (!tol_brent_p) tol_brent = 2.0e-4; else tol_brent = *tol_brent_p;
- if (!itmax_brent_p) itmax_brent = 100; else itmax_brent = *itmax_brent_p;
-
- fp=(*func)(p, n);
- (*dfunc)(xi, p, n, func, grdh_p, fp);
- for (j=n-1;j>=0;j--) {
- g[j] = -xi[j];
- xi[j]=h[j]=g[j];
- }
- memcpy(numgrad, xi, n*sizeof(double)); //Added by TZ, September 2003. Save the numerical gradient to be printed out at the right place.
- for (its=0;its<itmax;its++) {
- #ifdef PRINTON
- time(&begtime); //Added by TZ, September 2003.
- memcpy(pold, p, n*sizeof(double)); //Added by TZ, September 2003.
- #endif
- //====== Added by TZ, September 2003 ======
- if ( !(fptr_interesults = fopen(filename_sp3vecs,"w")) ) {
- printf("\n\nUnable to create the starting point data file %s in congradmin.c!\n", filename_sp3vecs);
- getchar();
- exit(EXIT_FAILURE);
- }
-// rewind(fptr_interesults); //Must put the pointer at the beginning of the file.
- //=== Prints out the line search direction.
- fprintf(fptr_interesults, "--------Line search direction---------\n");
- for (j=0; j<n; j++) fprintf(fptr_interesults, " %0.16e ", xi[j]);
- fprintf(fptr_interesults, "\n");
-// fflush( fptr_interesults );
-
-
- *iter=its;
- #if defined (CGI_OPTIMIZATION)
- linmin(p,xi,n,fret, tol_brent, itmax_brent, func);
- #elif defined (CGII_OPTIMIZATION)
- dlinmin(p, xi, n, fret, tol_brent, itmax_brent, grdh_p, func, dfunc);
- #else
- fn_DisplayError("The minimization routine frprmn() requires activating CGI_OPTIMIZATION or CGII_OPTIMIZATION in tzmatlab.h");
- #endif
- #ifdef PRINTON
- normforp = ftd_norm2(p, pold, n);
- //=== Prints out intermediate results.
- printf("\n========================================\n");
- printf("Intermediate results for the conjugate gradient algorithm.");
- printf("\n (1) Number of iterations so far (maximum number): %d (%d)\n (2) New value of log posterior kernal (old value, improvement): %0.9f (%0.9f, %0.9f)\n"
- " (3) Norm-2 of dx: %0.9f\n",
- its, itmax, -(*fret), -fp, fp-(*fret), normforp);
- fflush(stdout); // Flush the buffer to get out this message without delay.
- #endif
- //====== The following statements print out intermediate results. Added by TZ, September 2003 ======
- //=== Prints out the gradient.
- fprintf(fptr_interesults, "--------Numerical graident---------\n");
- for (j=0; j<n; j++) fprintf(fptr_interesults, " %0.16e ", numgrad[j]);
- fprintf(fptr_interesults, "\n");
- //
- fprintf(fptr_interesults, "--------Restarting point---------\n");
- for (j=0; j<n; j++) fprintf(fptr_interesults, " %0.16e ", p[j]);
- fprintf(fptr_interesults, "\n\n");
-// fflush( fptr_interesults );
- tzFclose(fptr_interesults);
-
-
- if (2.0*fabs(*fret-fp) <= ftol*(fabs(*fret)+fabs(fp)+EPS)) {
- //This is a normal convergence.
- printf("\n----- Normal convergence by the criterion of the objective function evaluation -----------\n");
- FREEALL
- return;
- }
- fp=(*func)(p, n);
- (*dfunc)(xi, p, n, func, grdh_p, fp);
- memcpy(numgrad, xi, n*sizeof(double)); //Added by TZ, September 2003. Save the numerical gradient to be printed out at the right place.
-// if (filename_sp3vecs) {
-// //=== Prints out the gradient.
-// fprintf(fptr_interesults, "--------Numerical graident---------\n");
-// for (j=0; j<n; j++) fprintf(fptr_interesults, " %0.16e ", xi[j]);
-// fprintf(fptr_interesults, "\n\n");
-//// fflush( fptr_interesults );
-
-// tzFclose(fptr_interesults);
-// }
- dgg=gg=0.0;
- for (j=n-1;j>=0;j--) {
- gg += g[j]*g[j];
- dgg += (xi[j]+g[j])*xi[j];
- }
- if (gg == 0.0) {
- FREEALL
- return;
- }
- gam=dgg/gg;
- for (j=n-1;j>=0;j--) {
- g[j] = -xi[j];
- xi[j]=h[j]=g[j]+gam*h[j];
- }
-
- #ifdef PRINTON
- time(¤time);
- //=== Times the iterative progress.
- printf(" (4) Seconds to complete one iteration: %0.4f\n (5) Current time of day: %s\n", difftime(currentime, begtime), ctime(¤time));
- fflush(stdout); // Flush the buffer to get out this message without delay.
- #endif
- }
- fn_DisplayError("The maximum number of iterations in frprmn() is reached before convergence");
-}
-#undef PRINTON
-#undef EPS
-#undef FREEALL
-
-
-#if defined (CGI_OPTIMIZATION)
- static int ncom;
- static double *pcom=NULL, *xicom=NULL, (*nrfunc)(double [], int); //nrfunc(), pcom, ncom, and xicom will be used by f1dim().
- static void linmin(double p[], double xi[], int n, double *fret, double tol_brent, int itmax_brent, double (*func)(double [], int)) {
- //Outputs:
- // p[0, ..., n-1]: a returned and reset value.
- // xi[0, ..., n-1]: a value repaced by the actual vector displacement that p was moved.
- // fret: the value of func at the returned location p.
- //Inputs:
- // p[0, ..., n-1]: a given point.
- // xi[0, ..., n-1]: a given multidimensional direction.
- // n: the dimension of p and xi.
- // func(): the objective function.
- int j;
- double xx,xmin,fx,fb,fa,bx,ax;
-
- ncom=n;
- pcom = tzMalloc(n, double);
- xicom = tzMalloc(n, double);
- nrfunc=func;
- for (j=n-1;j>=0;j--) {
- pcom[j]=p[j];
- xicom[j]=xi[j];
- }
- ax=0.0;
- xx=1.0;
- mnbrak(&ax,&xx,&bx,&fa,&fx,&fb,f1dim);
- *fret=brent(ax,xx,bx,f1dim, tol_brent, itmax_brent, &xmin);
- for (j=n-1;j>=0;j--) {
- xi[j] *= xmin;
- p[j] += xi[j];
- }
- tzDestroy(xicom);
- tzDestroy(pcom);
- }
-
-
- //=== Used by linmin() only;
- #define CGOLD 0.3819660
- #define ZEPS 1.0e-10
- #define SHFT(a,b,c,d) {(a)=(b);(b)=(c);(c)=(d);}
- #define SIGN(a,b) ((b) >= 0.0 ? fabs(a) : -fabs(a))
- static double brent(double ax, double bx, double cx, double (*f)(double), double tol_brent, double itmax_brent, double *xmin) {
- int iter;
- double a,b,d,etemp,fu,fv,fw,fx,p,q,r,tol1,tol2,u,v,w,x,xm;
- double e=0.0;
-
- a=(ax < cx ? ax : cx);
- b=(ax > cx ? ax : cx);
- x=w=v=bx;
- fw=fv=fx=(*f)(x);
- for (iter=0;iter<itmax_brent;iter++) {
- xm=0.5*(a+b);
- tol2=2.0*(tol1=tol_brent*fabs(x)+ZEPS);
- if (fabs(x-xm) <= (tol2-0.5*(b-a))) {
- *xmin=x;
- return fx;
- }
- if (fabs(e) > tol1) {
- r=(x-w)*(fx-fv);
- q=(x-v)*(fx-fw);
- p=(x-v)*q-(x-w)*r;
- q=2.0*(q-r);
- if (q > 0.0) p = -p;
- q=fabs(q);
- etemp=e;
- e=d;
- if (fabs(p) >= fabs(0.5*q*etemp) || p <= q*(a-x) || p >= q*(b-x))
- d=CGOLD*(e=(x >= xm ? a-x : b-x));
- else {
- d=p/q;
- u=x+d;
- if (u-a < tol2 || b-u < tol2)
- d=SIGN(tol1,xm-x);
- }
- } else {
- d=CGOLD*(e=(x >= xm ? a-x : b-x));
- }
- u=(fabs(d) >= tol1 ? x+d : x+SIGN(tol1,d));
- fu=(*f)(u);
- if (fu <= fx) {
- if (u >= x) a=x; else b=x;
- SHFT(v,w,x,u)
- SHFT(fv,fw,fx,fu)
- } else {
- if (u < x) a=u; else b=u;
- if (fu <= fw || w == x) {
- v=w;
- w=u;
- fv=fw;
- fw=fu;
- } else if (fu <= fv || v == x || v == w) {
- v=u;
- fv=fu;
- }
- }
- }
- fn_DisplayError("The maximum number of iterations in brent() is reached before convergence");
- *xmin=x;
- return fx;
- }
- #undef CGOLD
- #undef ZEPS
- #undef SHFT
- #undef SIGN
-
-#else //Default to CGII_OPTIMIZATION
-
- static int ncom;
- static double *pcom=NULL, *xicom=NULL, (*nrfunc)(double [], int); //nrfunc(), pcom, ncom, and xicom will be used by f1dim() and df1dim().
- static void (*nrdfun)(double [], double [], int, double (*func)(double [], int), double *, double);
- static void dlinmin(double p[], double xi[], int n, double *fret, double tol_dbrent, double itmax_dbrent, double *grdh_p, double (*func)(double [], int), void (*dfunc)(double [], double [], int, double (*func)(double [], int), double *, double)) {
- //Outputs:
- // p[0, ..., n-1]: a returned and reset value.
- // xi[0, ..., n-1]: a value repaced by the actual vector displacement that p was moved.
- // fret: the value of func at the returned location p.
- //Inputs:
- // p[0, ..., n-1]: a given point.
- // xi[0, ..., n-1]: a given multidimensional direction.
- // n: the dimension of p and xi.
- // func(): the objective function.
- // dfunc(): the gradient function computing the numerical gradient. In the form of gradcd_gen() in cstz.c.
-
- int j;
- double xx,xmin,fx,fb,fa,bx,ax;
-
- ncom=n;
- pcom = tzMalloc(n, double);
- xicom = tzMalloc(n, double);
- nrfunc=func;
- nrdfun=dfunc;
- for (j=n-1;j>=0;j--) {
- pcom[j]=p[j];
- xicom[j]=xi[j];
- }
- ax=0.0;
- xx=1.0;
- mnbrak(&ax,&xx,&bx,&fa,&fx,&fb,f1dim);
- *fret=dbrent(ax,xx,bx,f1dim, df1dim, grdh_p, tol_dbrent, itmax_dbrent, &xmin);
- for (j=n-1;j>=0;j--) {
- xi[j] *= xmin;
- p[j] += xi[j];
- }
- tzDestroy(xicom);
- tzDestroy(pcom);
- }
-
-
- //=== Used by dlinmin() only;
- #define ZEPS 1.0e-10
- #define MOV3(a,b,c, d,e,f) {(a)=(d);(b)=(e);(c)=(f);}
- #define SIGN(a,b) ((b) >= 0.0 ? fabs(a) : -fabs(a))
- static double dbrent(double ax, double bx, double cx, double (*f)(double), double (*df)(double, double *), double *grdh_p, double tol_dbrent, double itmax_dbrent, double *xmin) {
- int iter,ok1,ok2;
- double a,b,d,d1,d2,du,dv,dw,dx,e=0.0;
- double fu,fv,fw,fx,olde,tol1,tol2,u,u1,u2,v,w,x,xm;
-
- a=(ax < cx ? ax : cx);
- b=(ax > cx ? ax : cx);
- x=w=v=bx;
- fw=fv=fx=(*f)(x);
- dw=dv=dx=(*df)(x, grdh_p);
- for (iter=1;iter<=itmax_dbrent;iter++) {
- xm=0.5*(a+b);
- tol1=tol_dbrent*fabs(x)+ZEPS;
- tol2=2.0*tol1;
- if (fabs(x-xm) <= (tol2-0.5*(b-a))) {
- *xmin=x;
- return fx;
- }
- if (fabs(e) > tol1) {
- d1=2.0*(b-a);
- d2=d1;
- if (dw != dx) d1=(w-x)*dx/(dx-dw);
- if (dv != dx) d2=(v-x)*dx/(dx-dv);
- u1=x+d1;
- u2=x+d2;
- ok1 = (a-u1)*(u1-b) > 0.0 && dx*d1 <= 0.0;
- ok2 = (a-u2)*(u2-b) > 0.0 && dx*d2 <= 0.0;
- olde=e;
- e=d;
- if (ok1 || ok2) {
- if (ok1 && ok2)
- d=(fabs(d1) < fabs(d2) ? d1 : d2);
- else if (ok1)
- d=d1;
- else
- d=d2;
- if (fabs(d) <= fabs(0.5*olde)) {
- u=x+d;
- if (u-a < tol2 || b-u < tol2)
- d=SIGN(tol1,xm-x);
- } else {
- d=0.5*(e=(dx >= 0.0 ? a-x : b-x));
- }
- } else {
- d=0.5*(e=(dx >= 0.0 ? a-x : b-x));
- }
- } else {
- d=0.5*(e=(dx >= 0.0 ? a-x : b-x));
- }
- if (fabs(d) >= tol1) {
- u=x+d;
- fu=(*f)(u);
- } else {
- u=x+SIGN(tol1,d);
- fu=(*f)(u);
- if (fu > fx) {
- *xmin=x;
- return fx;
- }
- }
- du=(*df)(u, grdh_p);
- if (fu <= fx) {
- if (u >= x) a=x; else b=x;
- MOV3(v,fv,dv, w,fw,dw)
- MOV3(w,fw,dw, x,fx,dx)
- MOV3(x,fx,dx, u,fu,du)
- } else {
- if (u < x) a=u; else b=u;
- if (fu <= fw || w == x) {
- MOV3(v,fv,dv, w,fw,dw)
- MOV3(w,fw,dw, u,fu,du)
- } else if (fu < fv || v == x || v == w) {
- MOV3(v,fv,dv, u,fu,du)
- }
- }
- }
- fn_DisplayError("The maximum number of iterations in dbrent() is reached before convergence");
- return 0.0;
- }
- #undef ZEPS
- #undef MOV3
- #undef SIGN
-
- //=== Used by dlinmin() and dbrent() only;
- static double df1dim(double x, double *grdh_p) {
- int j;
- double df1=0.0;
- double *xt,*df;
-
- xt = tzMalloc(ncom, double);
- df = tzMalloc(ncom, double);
- for (j=ncom-1;j>=0;j--) xt[j]=pcom[j]+x*xicom[j];
- (*nrdfun)(df, xt, ncom, nrfunc, grdh_p, nrfunc(xt, ncom));
- //=================== WARNING ======================
- //We use 0.0 because the current gradient function gradcd_gen() in cstz.c do not use this function value. A more
- // sophisticated central gradient method would require this function value, and therefore we must pass
- // nrfunc(xt, ncom) instead of 0.0. TZ, September 2003.
- //=================== WARNING ======================
- for (j=ncom-1;j>=0;j--) df1 += df[j]*xicom[j];
- tzDestroy(df);
- tzDestroy(xt);
- return df1;
- }
-
-#endif
-
-
-
-static double f1dim(double x) {
- //Collapsing to one dimension line search, used by limin() or dlimin().
- int j;
- double f,*xt=NULL;
-
- xt = tzMalloc(ncom, double);
- for (j=ncom-1;j>=0;j--) xt[j]=pcom[j]+x*xicom[j];
- f=(*nrfunc)(xt, ncom);
- tzDestroy(xt);
- return f;
-}
-
-
-#define GOLD 1.618034
-#define GLIMIT 100.0
-#define TINY 1.0e-20
-#define SHFT(a,b,c,d) {(a)=(b);(b)=(c);(c)=(d);}
-#define SIGN(a,b) ((b) >= 0.0 ? fabs(a) : -fabs(a))
-static void mnbrak(double *ax, double *bx, double *cx, double *fa, double *fb, double *fc, double (*func)(double)) {
- double ulim,u,r,q,fu,dum, tmpd;
-
- *fa=(*func)(*ax);
- *fb=(*func)(*bx);
- if (*fb > *fa) {
- SHFT(dum,*ax,*bx,dum)
- SHFT(dum,*fb,*fa,dum)
- }
- *cx=(*bx)+GOLD*(*bx-*ax);
- *fc=(*func)(*cx);
- while (*fb > *fc) {
- r=(*bx-*ax)*(*fb-*fc);
- q=(*bx-*cx)*(*fb-*fa);
- u=(*bx)-((*bx-*cx)*q-(*bx-*ax)*r)/
- (2.0*SIGN((tmpd=fabs(q-r))>TINY ? tmpd : TINY,q-r)); //Original: (2.0*SIGN(FMAX(fabs(q-r),TINY),q-r));
- ulim=(*bx)+GLIMIT*(*cx-*bx);
- if ((*bx-u)*(u-*cx) > 0.0) {
- fu=(*func)(u);
- if (fu < *fc) {
- *ax=(*bx);
- *bx=u;
- *fa=(*fb);
- *fb=fu;
- return;
- } else if (fu > *fb) {
- *cx=u;
- *fc=fu;
- return;
- }
- u=(*cx)+GOLD*(*cx-*bx);
- fu=(*func)(u);
- } else if ((*cx-u)*(u-ulim) > 0.0) {
- fu=(*func)(u);
- if (fu < *fc) {
- SHFT(*bx,*cx,u,*cx+GOLD*(*cx-*bx))
- SHFT(*fb,*fc,fu,(*func)(u))
- }
- } else if ((u-ulim)*(ulim-*cx) >= 0.0) {
- u=ulim;
- fu=(*func)(u);
- } else {
- u=(*cx)+GOLD*(*cx-*bx);
- fu=(*func)(u);
- }
- SHFT(*ax,*bx,*cx,u)
- SHFT(*fa,*fb,*fc,fu)
- }
-}
-#undef GOLD
-#undef GLIMIT
-#undef TINY
-#undef SHFT
-#undef SIGN
-
-
-
-
-
-//-------------------
-// My own functions.
-//-------------------
-//=== Computing Norm2 of dx.
-static double ftd_norm2(double *vnew_p, double *vold_p, int _n) {
- int _i;
- double dtheta=0.0, //Cumulative.
- tmpd;
-
- for (_i=_n; _i>=0; _i--) {
- tmpd = vnew_p[_i] - vold_p[_i];
- dtheta += square(tmpd);
- }
-
- return ( sqrt(dtheta) );
-}
-
-
-
-//=== Extern function to be accessed by other C files.
-void congradmin_SetPrintFile(char *filename) {
- if (!filename) sprintf(filename_sp3vecs, "outdata5congradmin.prn"); //Default filename.
- else {
- strcpy(filename_sp3vecs, filename);
- //filename_sp3vecs[STRLEN-1] = '\0'; //The end of the string is set to NUL to prevent it from be a non-string.
- }
-}
-
-
-
-//void congradmin_SetPrintFile(FILE *fptr_sp) {
-// fptr_interesults = fptr_sp;
-//}
-
-//void congradmin_SetPrintFile_db(FILE *fptr_sp) {
-// fptr_interesults_db = fptr_sp;
-//}
-
-
-#undef STRLEN
diff --git a/CFiles/congradminOrigWorks.h b/CFiles/congradminOrigWorks.h
deleted file mode 100755
index 567c40aea4fad6287d6608d25812767cb005722e..0000000000000000000000000000000000000000
--- a/CFiles/congradminOrigWorks.h
+++ /dev/null
@@ -1,33 +0,0 @@
-#ifndef __CONGRADMIN_H__
-#define __CONGRADMIN_H__
-
- #include <math.h>
- #include <string.h>
-
- #include "tzmatlab.h"
-
-
- void frprmn(double p[], int n, int *iter, double *fret,
- double (*func)(double [], int), void (*dfunc)(double [], double [], int, double (*func)(double [], int), double *, double),
- double *ftol_p, int *itmax_p, double *tol_brent_p, int *itmax_brent_p, double *grdh_p);
- //Outputs:
- // p[0, ..., n-1]: the location of the minimum if it converges, which replaces the starting value.
- // iter: pointer to the number of iterations that were performed.
- // fret: pointer to the minimum value of the function.
- //Inputs:
- // p[0, ..., n-1]: a starting point for the minimization.
- // n: the dimension of p.
- // ftol_p: pointer to the convergence tolerance on the objective function value. Default: 1.0e-4 if NULL.
- // itmax_p: pointer to the maximum number of iterations in the main minimization program frprmn(). Default: 2000 if NULL.
- // tol_brent_p: pointer to the convergence tolerance for the line minimization in brent(). Default: 2.0e-4 if NULL.
- // itmax_brent_p: pointer to the maximum number of iterations for the line minimization in brent(). Default: 100 if NULL.
- // grdh: pointer to the user's specified step size for a numerical gradient. If NULL, dfunc() (i.e., gradcd_gen()) will select grdh automatically.
- // func(): the objective function.
- // dfunc(): the gradient function computing the numerical gradient. In the form of gradcd_gen() in cstz.c.
-
- void congradmin_SetPrintFile(char *filename);
- //If filename=NULL, no intermediate results will be printed out to a file.
-// void congradmin_SetPrintFile(FILE *fptr_sp);
- //If fptr_sp=NULL, no intermediate results will be printed out to a file.
-// void congradmin_SetPrintFile_db(FILE *fptr_sp);
-#endif
diff --git a/CFiles/csminwel.c b/CFiles/csminwel.c
index 6bae1803108d08100d1a814bdf11bc845cada849..1b3597d9936b2dacb27f10828c627f30ccbc9fc0 100755
--- a/CFiles/csminwel.c
+++ b/CFiles/csminwel.c
@@ -106,7 +106,7 @@ static char filename_sp2vecs[STRLEN]; //Two vectors. 1st row: numerical gradie
-#define MAX_NUM_BADCASES 3
+#define MAX_NUM_POSSIBLE_BADCASES 250 //After this number, reset inverse hessian (especially dealing with the case for "Correct for low angle").
#define EPS (1.0e-10) //Small number to rectify special case of converging to exactly zero function value.
#define TERMINATEVALUE (1.0e+300) //If the value of the objective function at the intial value is greater than this, terminates the program.
void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
@@ -205,7 +205,7 @@ void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
/* if stuck, give it another try by perturbing Hessian */
memcpy(Hcliff,H,nn*sizeof(double));
for (i=0; i<nn; i+=n+1)
- Hcliff[i] *= 1+rand()*randmax; //DDDDebugging. Hcliff[i] *= 1+0.5;
+ Hcliff[i] *= 1+rand()*randmax; //Random search. DDDDebugging. Hcliff[i] *= 1+0.5;
#ifdef VERBOSE_WARNINGS
printf("======= Random search takes place now. =======\n");
@@ -372,13 +372,16 @@ void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
if ((int)*itct > nit) {
#ifdef VERBOSE_WARNINGS
- printf("\nWarning: termination as the maximum number of iterations is reached.\n");
+ printf("\nWarning: Termination as the maximum number of iterations %d is reached with return code %d.\n", nit, *retcodeh);
+ fprintf(FPTR_DEBUG, "\nWarning: Termination as the maximum number of iterations %d is reached with return code %d.\n", nit, *retcodeh);
#endif
done = 1;
}
- else if (stuck) {
+ else if (stuck) //stuck = 1 means fabs(*fh-f[0]) < crit, which means converged. why do we use the word "stuck" is beyond me.
+ {
#ifdef VERBOSE_DETOUTPUT
printf("\nConvergence (improvement < crit %.4e) with return code %d.\n", crit, *retcodeh);
+ fprintf(FPTR_DEBUG, "\nConvergence (improvement < crit %.4e) with return code %d.\n", crit, *retcodeh);
#endif
done = 1;
@@ -388,7 +391,7 @@ void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
#ifdef VERBOSE_WARNINGS
switch (*retcodeh) {
case 1:
- printf("\nCoverged: Zero gradient.\n"); break;
+ printf("\nConverged: Zero gradient.\n"); break;
case 2:
printf("\nWarning: Back adjustment of stepsize didn't finish.\n"); break;
case 3:
@@ -404,16 +407,19 @@ void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
}
#endif
- //=== Restarts from the initial (inverse of) Hessian when stuck for a while in bad cases. Added by TZ.
- if (*retcodeh && *retcodeh != 1)
- if (++cnt_n_badcases >= MAX_NUM_BADCASES) {
+ //=== Restarts from the initial (inverse of) Hessian when stuck for a while in bad cases. This can happen even *retcodeh=0 while stuck=0.
+ //if (*retcodeh && *retcodeh != 1)
+ if (!done)
+ if (cnt_n_badcases++ >= MAX_NUM_POSSIBLE_BADCASES)
+ {
H_dm->M = H;
H_dm->nrows = H_dm->ncols = n;
InitializeDiagonalMatrix_lf(H_dm, GLB_sclForHess);
//H_dm->flag = M_GE | M_SU | M_SL; //Hessian is symmetric.
cnt_n_badcases = 0; //Reset after we restart wtih the initial Hessian.
#ifdef VERBOSE_WARNINGS
- printf("Hessian is reset to the initial value because the maximum number of bad cases, %d, is reached!\n", MAX_NUM_BADCASES);
+ printf("Hessian is reset to the initial value because the maximum number of possibly bad cases, %d, is reached with return code %d!\n", MAX_NUM_POSSIBLE_BADCASES, *retcodeh);
+ fprintf(FPTR_DEBUG, "Hessian is reset to the initial value because the maximum number of possibly bad cases, %d, is reached with return code %d!\n", MAX_NUM_POSSIBLE_BADCASES, *retcodeh);
#endif
}
@@ -448,7 +454,7 @@ void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
}
tzDestroy(H_dm);
}
-#undef MAX_NUM_BADCASES
+#undef MAX_NUM_POSSIBLE_BADCASES
#undef EPS
#undef TERMINATEVALUE
diff --git a/CFiles/csminwel.h b/CFiles/csminwel.h
old mode 100755
new mode 100644
diff --git a/CFiles/csminwelOrigWorks.c b/CFiles/csminwelOrigWorks.c
deleted file mode 100755
index fc118aa7acfe4b99451ddd3642f1903ea5d6609c..0000000000000000000000000000000000000000
--- a/CFiles/csminwelOrigWorks.c
+++ /dev/null
@@ -1,706 +0,0 @@
-/*=========================================================
- * csminwel.c
- *
- * Unconstrained minimization. Uses a quasi-Newton method with BFGS update of
- * the estimated inverse hessian. It is robust against certain pathologies
- * common on likelihood functions. It attempts to be robust against "cliffs",
- * i.e. hyperplane discontinuities, though it is not really clear whether what
- * it does in such cases succeeds reliably.
- *
- * function [fhat,xhat,ghat,Hhat,itct,fcount,retcodehat] = csminwelmex(fcn,x0,H0,grad,crit,nit,varargin)
- * fcn: string naming the objective function to be minimized
- * x0: initial value of the parameter vector
- * H0: initial value for the inverse Hessian. Must be positive definite.
- * grad: Either a string naming a function that calculates the gradient, or the null matrix.
- * If it's null, the program calculates a numerical gradient. In this case fcn must
- * be written so that it can take a matrix argument and produce a row vector of values.
- * crit: Convergence criterion. Iteration will cease when it proves impossible to improve the
- * function value by more than crit.
- * nit: Maximum number of iterations.
- * varargin: A list of optional length of additional parameters that get handed off to fcn each
- * time it is called.
- * Note that if the program ends abnormally, it is possible to retrieve the current x,
- * f, and H from the files g1.mat and H.mat that are written at each iteration and at each
- * hessian update, respectively. (When the routine hits certain kinds of difficulty, it
- * write g2.mat and g3.mat as well. If all were written at about the same time, any of them
- * may be a decent starting point. One can also start from the one with best function value.)
- *
- * Note: to set the level of display output, change preprocessor definitions in csminwel.h
- * to display all output, uncomment both VERBOSE_WARNINGS and VERBOSE_DETOUTPUT
- * to display only warnings without output, uncomment VERBOSE_WARNINGS
- * to display no ouput, comment both VERBOSE_DETOUTPUT and VERBOSE_WARNINGS
- *
- * MATLAB algorithm by Christopher Sims
- * C implementation by Iskander Karibzhanov
- * Modified by Dan Waggoner and Tao Zha
- *
- * Copyright(c) 1996 Christopher Sims
- * Copyright(c) 2003 Karibzhanov, Waggoner, and Zha
- *=======================================================
- * Revision history:
- *
- * 10/3/2002 - 1. corrected problem with memory corruption in C-MEX-file (csminwelmex.c)
- * (needed to switch fcnRhs[0] back to x[0] before destroying it.
- * If we don't do this, we will later clear previously destroyed array
- * (one of x[1], x[2] or x[3]) which causes memory fault.
- * The reason why fcnRhs[0] pointed to array other than x[0] is
- * because we use mxSetPr in feval and gfeval.
- * This was not a problem in C-file (csminwel.c).
- *
- * 10/11/2002 - 1. changed csminit function to avoid using fPeak without first initializing it
- * 2. added two switches in csminit function to assign retcode to 7 for lambda>=4
- * 3. added one more verbose level to display only warnings or all output *
- *=======================================================*/
-
-#include "csminwel.h"
-#define STRLEN 192
-static int numgrad(double *g, double *x, int n,
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims);
-static void csminit(double *fhat, double *xhat, int *fcount, int *retcode,
- double *x0, double f0, double *g, int badg, double *H0, int n,
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims);
-static void bfgsi(double *H, double *dg, double *dx, int n, int nn);
-static int peakwall(double *g, int retcode, double *x, int n,
- int (*gfcn)(double *x, int n, double *g, double **args, int *dims),
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims);
-static double times(double *x, double *y, int n);
-static double *mtimes(double *x, double *y, int n, int nn);
-static double *mminus(double *x, double *y, int n);
-
-
-static FILE *fptr_interesults = (FILE *)NULL; //Printing intermediate results to a file.
-static char filename_sp2vecs[STRLEN]; //Two vectors. 1st row: numerical gradient; 2nd row: vectorized parameters.
-
-
-void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
- double *xh, int n, double *H, double *gh,
- int (*gfcn)(double *x, int n, double *g, double **args,
- int *dims), double *fh, double crit, int *itct, int nit,
- int *fcount, int *retcodeh, double **args, int *dims) {
-
- int done=0, badg[4], badgh, nogh=1, stuck=0;
- double *x[4], *g[4], f[4], *dg, *dx;
- int retcode[3], fc=0, ih, nn, i;
- #ifdef VERBOSE_DETOUTPUT
- time_t begtime, currentime;
- #endif
-
-
- nn = n*n; /* n: dimension size of x or xh */
- *itct = -1; /* itct: number of actual iterations */
- *fcount = -1; /* fcount: number of evaluations of the function */
-
- for (i=0; i<4; i++)
- x[i] = tzMalloc(n, double); //x[i] = calloc(n, sizeof(double)); Commented out by TZ.
- memcpy(x[0],xh,n*sizeof(double));
-
- for (i=0; i<4; i++)
- g[i] = tzMalloc(n, double); //calloc(n, sizeof(double)); Commented out by TZ.
-
- f[0] = fcn(x[0],n,args,dims);
-
- if (f[0] > 1.0e+50) {
- printf("Bad initial parameter.\n");
- return;
- }
-
- if (gfcn)
- /* if grad is a string, compute it */
- badg[0] = gfcn(x[0],n,g[0],args,dims);
- else
- /* if grad is not string, compute it */
- badg[0] = numgrad(g[0],x[0],n,fcn,args,dims);
- retcode[2] = 101;
- /* iterate until done is false */
- while (!done) {
- #ifdef VERBOSE_DETOUTPUT
- time(&begtime);
- #endif
-
- for (i=0; i<n; i++)
- g[1][i] = g[2][i] = g[3][i] = 0;
-
-// #ifdef VERBOSE_DETOUTPUT
-// printf("-----------------\n-----------------\n");
-// printf("f at the beginning of new iteration, %.10f\nx = ",f[0]);
-// for (i=0; i<n; i++) {
-// printf("%15.8g ",x[0][i]);
-// if (i%4==3) printf("\n");
-// }
-// if (i%4>0) printf("\n");
-// #endif
-
- (*itct)++;
- csminit(&f[1],x[1],&fc,&retcode[0],x[0],f[0],g[0],badg[0],H,n,fcn,args,dims);
- *fcount += fc;
- /* if retcode1=1 gradient is zero and you are at the peak */
- if (retcode[0]!=1) {
- badg[1] = peakwall(g[1],retcode[0],x[1],n,gfcn,fcn,args,dims);
- /* Bad gradient or back and forth on step length.
- Possibly at cliff edge. Try perturbing search direction. */
- if (badg[1]) {
- double *Hcliff = tzMalloc(nn, double); //calloc(nn,sizeof(double)); Commented out by TZ.
- double randmax=1/RAND_MAX;
- /* if stuck, give it another try by perturbing Hessian */
- memcpy(Hcliff,H,nn*sizeof(double));
- for (i=0; i<nn; i+=n+1)
- Hcliff[i] *= 1+rand()*randmax;
-
- #ifdef VERBOSE_WARNINGS
- printf("Cliff. Perturbing search direction.\n");
- #endif
-
- csminit(&f[2],x[2],&fc,&retcode[1],x[0],f[0],g[0],badg[0],Hcliff,n,fcn,args,dims);
- *fcount += fc;
- if (f[2] < f[0]) {
- badg[2] = peakwall(g[2],retcode[1],x[2],n,gfcn,fcn,args,dims);
- if (badg[2]) {
- double *xx = tzMalloc(n, double), nx; //calloc(n,sizeof(double)), nx; Commented out by TZ.
-
- #ifdef VERBOSE_WARNINGS
- printf("Cliff again. Try traversing.\n");
- #endif
-
- for (i=0; i<n; i++)
- xx[i] = x[2][i]-x[1][i];
- nx = times(xx,xx,n);
- if (sqrt(nx) < 1e-13) {
- f[3] = f[0];
- memcpy(x[3],x[0],n*sizeof(double));
- badg[3] = 1;
- retcode[2] = 101;
- } else {
- double *gcliff = tzMalloc(n, double), //calloc(n,sizeof(double)), Commented out by TZ.
- *eye = tzMalloc(n, double); //calloc(n,sizeof(double)); Commented out by TZ.
- double dfnx = (f[2]-f[1])/nx;
- for (i=0; i<n; i++) {
- gcliff[i] = dfnx*xx[i];
- eye[i*(n+1)] = 1;
- }
- csminit(&f[3],x[3],&fc,&retcode[2],x[0],f[0],gcliff,0,eye,n,fcn,args,dims);
- *fcount += fc;
- badg[3] = peakwall(g[3],retcode[2],x[3],n,gfcn,fcn,args,dims);
- free(eye);
- free(gcliff);
- }
- free(xx);
- } else {
- f[3] = f[0];
- memcpy(x[3],x[0],n*sizeof(double));
- badg[3] = 1;
- retcode[2] = 101;
- }
- } else {
- f[3] = f[0];
- memcpy(x[3],x[0],n*sizeof(double));
- badg[3] = 1;
- retcode[2] = 101;
- }
- free(Hcliff);
- } else {
- /* normal iteration, no walls, or else we're finished here. */
- f[2] = f[0];
- f[3] = f[0];
- badg[2] = 1;
- badg[3] = 1;
- retcode[1] = 101;
- retcode[2] = 101;
- }
- }
-
-
- /* how to pick gh and xh */
- if (f[3]<f[0] && badg[3]==0) {
- /* if 3 (transversing) was needed, it improved and gradient is good, take that */
- ih = 3;
- *fh = f[3];
- memcpy(xh,x[3],n*sizeof(double));
- memcpy(gh,g[3],n*sizeof(double));
- badgh = badg[3];
- *retcodeh = retcode[2];
- }
- else if (f[2]<f[0] && badg[2]==0) {
- /* if 2 (perturbig) was needed, it improved and gradient is good, take that */
- ih = 2;
- *fh = f[2];
- memcpy(xh,x[2],n*sizeof(double));
- memcpy(gh,g[2],n*sizeof(double));
- badgh = badg[2];
- *retcodeh = retcode[1];
- }
- else if (f[1]<f[0] && badg[1]==0) {
- /* if first try went fine, take that */
- ih = 1;
- *fh = f[1];
- memcpy(xh,x[1],n*sizeof(double));
- memcpy(gh,g[1],n*sizeof(double));
- badgh = badg[1];
- *retcodeh = retcode[0];
- }
- else {
- /* if nothing worked, just take the min of your attempts and compute the gradient */
- if (f[1] <= f[2])
- if (f[1] <= f[3]) ih = 1;
- else ih = 3;
- else if (f[2] <= f[3]) ih = 2;
- else ih = 3;
- *fh = f[i];
- memcpy(xh,x[ih],n*sizeof(double));
- *retcodeh = retcode[ih-1];
- if (nogh) {
- nogh = 0;
- if (gfcn)
- badgh = gfcn(xh,n,gh,args,dims);
- else
- badgh = numgrad(gh,xh,n,fcn,args,dims);
- }
- badgh = 1;
- }
- /* end of picking */
- stuck = fabs(*fh-f[0]) < crit;
- /* if nothing REALLY worked, too bad, you're stuck */
- if (!badg[0] && !badgh && !stuck) {
- /* if you are not stuck, update H0 matrix */
- dg = mminus(gh,g[0],n);
- dx = mminus(xh,x[0],n);
- bfgsi(H,dg,dx,n,nn);
- free(dx);
- free(dg);
- }
-
- #ifdef VERBOSE_DETOUTPUT
- //=== Prints out intermediate results.
- printf("========================================\n");
- printf(" (1) New value of the obj. func. on iteration %d: %.9f\n (2) Old value: %.9f\n (3) Downhill improvement: %.9f\n",
- (int)*itct, *fh, f[0], f[0]-(*fh));
-
- time(¤time);
- //=== Times the iterative progress.
- printf(" (4) Seconds to complete one iteration: %0.4f\n (5) Current time of day: %s\n\n", difftime(currentime, begtime), ctime(¤time));
- fflush(stdout); // Flush the buffer to get out this message without delay.
- #endif
-
- //--------- Prints outputs to a file. ---------
- if ( !(fptr_interesults = fopen(filename_sp2vecs,"w")) ) {
- printf("\n\nUnable to create the starting point data file %s in csminwel.c!\n", filename_sp2vecs);
- getchar();
- exit(EXIT_FAILURE);
- }
- fprintf(fptr_interesults, "--------Numerical gradient---------\n");
- for (i=0; i<n; i++) fprintf(fptr_interesults, " %0.16e ", gh[i]);
- fprintf(fptr_interesults, "\n");
- fprintf(fptr_interesults, "--------Restarting point---------\n");
- for (i=0; i<n; i++) fprintf(fptr_interesults, " %0.16e ", xh[i]);
- fprintf(fptr_interesults, "\n\n");
- tzFclose(fptr_interesults);
-
- if ((int)*itct > nit) {
- #ifdef VERBOSE_WARNINGS
- printf("\nWarning: termination as the maximum number of iterations is reached.\n");
- #endif
- done = 1;
- }
- else if (stuck) {
- #ifdef VERBOSE_DETOUTPUT
- printf("improvement < crit termination\n");
- #endif
-
- done = 1;
- }
-
- #ifdef VERBOSE_WARNINGS
- switch ((int)*retcodeh) {
- case 1:
- printf("\nWarning: Zero gradient.\n"); break;
- case 2:
- printf("\nWarning: Back adjustment of stepsize didn't finish.\n"); break;
- case 3:
- printf("\nWarning: Smallest stepsize still improving too slow.\n"); break;
- case 4:
- printf("\nWarning: Forth adjustment of stepsize didn't finish.\n"); break;
- case 6:
- printf("\nWarning: Smallest step still improving too slow, reversed gradient.\n"); break;
- case 5:
- printf("\nWarning: Largest stepsize still improving too fast.\n"); break;
- case 7:
- printf("\nWarning: Possible inaccuracy in Hessian matrix.\n"); break;
- }
- #endif
-
- f[0] = *fh;
- memcpy(x[0],xh,n*sizeof(double));
- memcpy(g[0],gh,n*sizeof(double));
- badg[0] = badgh;
- }
- for (i=0; i<4; i++) {
- free(g[i]);
- free(x[i]);
- }
-}
-
-/**/
-#define SCALE 1.0
-static int numgrad(double *g, double *x, int n,
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims) {
- //Forward difference gradient method.
- double delta, deltai;
- double f0, g0, ff, tmp, *xp;
- int i;
- int badg;
- f0 = fcn(x,n,args,dims);
- badg = 0;
- for (i=0, xp=x; i<n; i++, xp++, g++) {
- delta=SCALE*1.0e-5, deltai=1.0e+5/SCALE;
-
- tmp = *xp;
- *xp += delta;
- delta = *xp - tmp; // This increases the precision slightly. Added by TZ.
- if ( (ff=fcn(x,n,args,dims)) < NEARINFINITY ) g0 = (ff-f0)*deltai; //Not over the boundary.
- else {
- //Switches to the other side of the boundary.
- *xp = tmp - delta;
- g0 = (f0-fcn(x,n,args,dims))*deltai;
- }
-
- *xp = tmp; //Puts back to the original place. TZ, 9/03.
- if (fabs(g0)<1.0e+15)
- *g = g0;
- else {
- #ifdef VERBOSE_WARNINGS
- printf("Bad gradient.\n");
- #endif
-
- *g = 0;
- badg = 1;
- }
- }
- return badg;
-}
-#undef SCALE
-
-/**
-#define STPS 1.0e-04 // 6.0554544523933391e-6 step size = pow(DBL_EPSILON,1.0/3)
-static int numgrad(double *g, double *x, int n,
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims) {
- //Central difference gradient method. Added by TZ.
- double dh;
- double f0, fp, fm, tmp, *xp;
- int i;
- int badg;
-
- badg = 0;
- for (i=0, xp=x; i<n; i++, xp++, g++) {
- dh = fabs(*xp)<=1 ? STPS : STPS*(*xp);
-
- tmp = *xp;
- *xp += dh;
- dh = *xp - tmp; // This increases the precision slightly.
- fp = fcn(x,n,args,dims);
- *xp = tmp - dh;
- fm = fcn(x,n,args,dims);
-
- //=== Checking the boundary condition for the minimization problem.
- if (fp >= NEARINFINITY) {
- *xp = tmp; //Puts back to the original place. TZ, 9/03.
- f0 = fcn(x,n,args,dims);
- *g = (f0-fm)/dh;
- }
- else if (fm >= NEARINFINITY) {
- //Switches to the other side of the boundary.
- *xp = tmp; //Puts back to the original place. TZ, 9/03.
- f0 = fcn(x,n,args,dims);
- *g = (fp-f0)/dh;
- }
- else {
- *g = (fp-fm)/(2.0*dh);
- *xp = tmp; //Puts back to the original place. TZ, 9/03.
- }
-
- if (fabs(*g)>1.0e+15) {
- #ifdef VERBOSE_WARNINGS
- printf("Bad gradient.\n");
- #endif
- *g = 0.0;
- badg = 1;
- }
- }
- return badg;
-}
-#undef STPS
-/**/
-
-
-#define ANGLE 0.05 //When output of this variable becomes negative, we have a wrong analytical graident.
- //.005 works for identified VARs and OLS.
-#define THETA .4 //(0<THETA<.5) THETA near .5 makes long line searches, possibly fewer iterations.
- //.1 works for OLS or other nonlinear functions.
- //.3 works for identified VARs.
-#define FCHANGE 1000
-#define MINLAMB 1e-9
-#define MINDFAC .01
-static void csminit(double *fhat, double *xhat, int *fcount, int *retcode,
- double *x0, double f0, double *g, int badg, double *H0, int n,
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims) {
- double lambda=1, gnorm=0, dxnorm=0, factor=3, lambdaPeak=0;
- double f, dfhat, a, tmp, fPeak=f0, lambdaMax=DBL_MAX;
- double *dx, *dxtest;
- int done=0, shrink=1, shrinkSignal, growSignal;
- int i;
-
- *fhat = f0;
- *fcount = 0;
- *retcode = 0;
- gnorm = sqrt(times(g,g,n));
- if ((gnorm < 1.e-12) && !badg)
- *retcode = 1; /* gradient convergence */
- else {
- /* with badg 1, we don't try to match rate of improvement to directional
- derivative. We're satisfied just to get some improvement in f. */
- dx = tzMalloc(n, double); //dx = calloc(n, sizeof(double)); Commented out by TZ.
- //if (!dx) printf("Dynamic memory allocation error.\n"); Commnted out by TZ.
- for (i=0; i<n; i++)
- dx[i] = -times(&H0[i*n],g,n);
- dxnorm = sqrt(times(dx,dx,n));
- if (dxnorm > 1e12) {
- #ifdef VERBOSE_WARNINGS
- printf("Near-singular H problem.\n");
- #endif
-
- for (i=0; i<n; i++)
- dx[i] *= FCHANGE/dxnorm;
- }
- dfhat = times(dx,g,n);
- if (!badg) {
- /* test for alignment of dx with gradient and fix if necessary */
- a = -dfhat/(gnorm*dxnorm);
- if (a<ANGLE) {
- tmp = (ANGLE*dxnorm+dfhat/gnorm)/gnorm;
- for (i=0; i<n; i++)
- dx[i] -= tmp*g[i];
- dfhat = times(dx,g,n);
- dxnorm = sqrt(times(dx,dx,n));
-
- #ifdef VERBOSE_DETOUTPUT
- printf("Correct for low angle: %g\n",a);
- #endif
- }
- }
-
- #ifdef VERBOSE_DETOUTPUT
- printf("Predicted improvement: %18.9f, Norm of gradient: %18.9f\n",-dfhat/2,gnorm);
- #endif
-
- dxtest = tzMalloc(n, double); //calloc(n, sizeof(double)); Commented out by TZ.
- while (!done) {
- for (i=0; i<n; i++)
- dxtest[i] = x0[i]+dx[i]*lambda;
- f = fcn(dxtest,n,args,dims);
-
- #ifdef VERBOSE_DETOUTPUT
- printf("lambda = %10.5g; f = %20.7f\n",lambda,f);
- #endif
-
- if (f<*fhat) {
- *fhat = f;
- memcpy(xhat,dxtest,n*sizeof(double));
- }
- (*fcount)++;
- tmp = -THETA*dfhat*lambda;
-
- /* the optimal lambda should be such that f0-f > -THETA*dfhat*lambda (see Berndt et al.)
- If that's not the case and grad is good, OR
- if grad is bad and f is not going down, shrinkSignal = 1 */
- shrinkSignal = ( !badg && (f0-f <= (tmp>0?tmp:0)) ) ||
- ( badg && (f0-f < 0 ) );
-
- /* the optimal lambda should also be such that f0-f<-(1-THETA)*dfhat*lambda
- If that's not the case with lambda>0, AND grad is good, growthSignal = 1 */
- growSignal = !badg && ( (lambda > 0) && (f0-f >= -(1-THETA)*dfhat*lambda) );
-
- /* If shrinkSignal=1 AND ( lambda>lambdaPeak or lambda negative )
- (note when lambdaPeak=0 the second part only excludes lambda=0)
- try shrinking lambda */
- if ( shrinkSignal && ( (lambda>lambdaPeak) || (lambda<0) ) ) {
- /* if shrink=0 OR lambda/factor is already smaller than lambdaPeak, increase factor */
- if ( (lambda>0) && ((!shrink) || (lambda/factor <= lambdaPeak)) ) {
- shrink = 1;
- factor = pow(factor,.6);
- while (lambda/factor <= lambdaPeak)
- factor = pow(factor,.6);
- if (fabs(factor-1)<MINDFAC) {
- if (fabs(lambda) < 4)
- *retcode = 2;
- else
- *retcode = 7;
- done = 1;
- }
- }
- if ((lambda<lambdaMax) && (lambda>lambdaPeak))
- lambdaMax=lambda;
- /* shrink lambda */
- lambda /= factor;
- /* if lambda has already been shrunk as much as possible */
- if (fabs(lambda) < MINLAMB)
- /* if lambda is positive AND you have not made any improvement
- try going against gradient, which may be inaccurate */
- if ((lambda > 0) && (f0 <= *fhat))
- lambda = -lambda*pow(factor,6);
- else {
- /* if lambda is negative: let it be known and quit trying */
- if (lambda < 0)
- *retcode = 6;
- /* if you have not made any imporvement:
- let it be known and quit trying */
- else
- *retcode = 3;
- done = 1;
- }
- }
- /* If growSignal=1 and lambda positive OR ( lambda>lambdaPeak or lambda negative )
- (note when lambdaPeak=0 the second part only excludes lambda=0)
- try increase lambda */
- else
- if ( (growSignal && (lambda > 0) ) ||
- ( shrinkSignal && (lambda <= lambdaPeak) && (lambda > 0) ) ) {
- if (shrink) {
- shrink = 0;
- factor = pow(factor,.6);
- if (fabs(factor-1) < MINDFAC) {
- if (fabs(lambda) < 4)
- *retcode = 4;
- else
- *retcode = 7;
- done = 1;
- }
- }
- if ( (f<fPeak) && (lambda>0) ) {
- fPeak = f;
- lambdaPeak = lambda;
- if (lambdaMax <= lambdaPeak)
- lambdaMax = lambdaPeak*factor*factor;
- }
- /* increase lambda (up to 1e20) */
- lambda *= factor;
- /* if lambda has been increased up to the limit and
- you have not made any imporvement:
- let it be known and quit trying */
- if (fabs(lambda) > 1e20) {
- *retcode = 5;
- done = 1;
- }
- }
- /* If growthSignal=shrinkSignal=0 you found a good lambda, you are done */
- else {
- done = 1;
- *retcode = factor<1.2 ? 7 : 0;
- }
- }
- free(dxtest);
- free(dx);
- }
- #ifdef VERBOSE_DETOUTPUT
- printf("Norm of dx %10.5g\n", dxnorm);
- #endif
-}
-#undef ANGLE
-#undef THETA
-#undef FCHANGE
-#undef MINLAMB
-#undef MINDFAC
-
-
-static double times(double *x, double *y, int n) {
- double z = 0;
- int i;
- for (i=0; i<n; i++, x++, y++)
- z += (*x)*(*y);
- return z;
-}
-
-static int peakwall(double *g, int retcode, double *x, int n,
- int (*gfcn)(double *x, int n, double *g, double **args, int *dims),
- double (*fcn)(double *x, int n, double **args, int *dims),
- double **args, int *dims) {
- /* if retcode=2 or 4 you have shrunk or increased lambda as much as you could:
- exhausted search possibilities the csminit step has failed */
- if (retcode==2 || retcode==4)
- return 1;
- else
- /* if you are not at the peak but the csminit has improved,
- compute the gradient again to update H0 */
- if (gfcn)
- return gfcn(x,n,g,args,dims);
- else
- return numgrad(g,x,n,fcn,args,dims);
-}
-
-static void bfgsi(double *H, double *dg, double *dx, int n, int nn) {
- double *Hdg, *dxdx, *dxHdg, *Hdgdx;
- double dgdx, m;
- int i;
-
- Hdg = tzMalloc(n, double); //calloc(n, sizeof(double)); Commented out by TZ.
- //if (!Hdg) printf("Dynamic memory allocation error.\n"); Commented out by TZ.
-
- /* Hdg = H0*dg; */
- for (i=0; i<n; i++)
- Hdg[i] = times(&H[i*n],dg,n);
- /* dgdx = dg'*dx; */
- dgdx = 1/times(dg,dx,n);
- if (fabs(dgdx)<1e12) {
- dxdx = mtimes(dx,dx,n,nn);
- dxHdg = mtimes(dx,Hdg,n,nn);
- Hdgdx = mtimes(Hdg,dx,n,nn);
- m = 1+times(dg,Hdg,n)*dgdx;
- for (i=0; i<nn; i++, H++, dxdx++, dxHdg++, Hdgdx++)
- *H += (m*(*dxdx)-(*dxHdg)-(*Hdgdx))*dgdx;
- free(Hdgdx-nn);
- free(dxHdg-nn);
- free(dxdx-nn);
- }
- else {
- #ifdef VERBOSE_WARNINGS
- printf("BFGS update failed.\n");
- printf("|dg| = %f |dx| = %f\n",sqrt(times(dg,dg,n)),sqrt(times(dx,dx,n)));
- printf("dg\'*dx = %f\n",dgdx);
- printf("|H*dg| = %f\n",sqrt(times(Hdg,Hdg,n)));
- #endif
- }
- free(Hdg);
-}
-
-static double *mtimes(double *x, double *y, int n, int nn) {
- double *x0;
- double *z;
- int i, j;
- z = tzMalloc(nn, double); //calloc(nn, sizeof(double)); Commented out by TZ.
- for (i=0, x0=x; i<n; i++, y++)
- for (j=0, x=x0; j<n; j++, x++, z++)
- *z = (*x)*(*y);
- return z-nn;
-}
-
-static double *mminus(double *x, double *y, int n) {
- double *z;
- int i;
- z = tzMalloc(n, double); //calloc(n, sizeof(double)); Commented out by TZ.
- for (i=0; i<n; i++, x++, y++, z++)
- *z = (*x)-(*y);
- return z-n;
-}
-
-
-//=== Extern function to be accessed by other C files.
-void csminwel_SetPrintFile(char *filename) {
- if (!filename) sprintf(filename_sp2vecs, "outdata5csminwel.prn"); //Default filename.
- else if (STRLEN-1 < strlen(filename)) fn_DisplayError(".../csminwel.c: the allocated length STRLEN for filename_sp2vecs is too short. Must increase the string length");
- else strcpy(filename_sp2vecs, filename);
-}
-
-#undef STRLEN
diff --git a/CFiles/csminwelOrigWorks.h b/CFiles/csminwelOrigWorks.h
deleted file mode 100755
index bb4e8daf5e8fe14e59f49d310003e5f0d43ae92c..0000000000000000000000000000000000000000
--- a/CFiles/csminwelOrigWorks.h
+++ /dev/null
@@ -1,22 +0,0 @@
-#ifndef __CSMINWEL_H__
-#define __CSMINWEL_H__
-
- #define VERBOSE_WARNINGS // display warnings
- #define VERBOSE_DETOUTPUT // display detailed output
-
- #include <math.h>
- #include <string.h>
- #include <stdio.h>
- #include <float.h>
-
- #include "tzmatlab.h"
-
- void csminwel(double (*fcn)(double *x, int n, double **args, int *dims),
- double *x, int n, double *H, double *gh,
- int (*grad)(double *x, int n, double *g, double **args,
- int *dims), double *fh, double crit, int *itct, int nit,
- int *fcount, int *retcodeh, double **args, int *dims);
- // Alternative but less clear way: ... (double (*fcn)(double *, int, double **, int *), ...
-
- void csminwel_SetPrintFile(char *filename);
-#endif
diff --git a/CFiles/cstz.c b/CFiles/cstz.c
old mode 100755
new mode 100644
index 5602ee686e203f0e0857c8cd01102d6ad508a4d1..553d6a5fc67a55961c0295bfa7c1875afb8f8da0
--- a/CFiles/cstz.c
+++ b/CFiles/cstz.c
@@ -1615,104 +1615,121 @@ double fn_replace_logofsumsbt(double *yold, double _a, double ynew, double _b)
}
-
-
-
//<<---------------
// Evaluating the inverse of the chi-square cumulative distribution function.
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double fn_chi2inv(double p, double df)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
//Returns x where p = int_{0}^{x} chi2pdf(t, df) dt
if (df<=0.0) fn_DisplayError("cstz.c/fn_chi2inv(): degrees of freedom df must be greater than 0.0");
if (p<=0.0) return (0.0);
else if (p>=1.0) return (MACHINEINFINITY);
else return (imsls_d_chi_squared_inverse_cdf(p, df));
-}
+#elif defined( USE_GSL_LIBRARY )
+ if (df<=0.0) fn_DisplayError("cstz.c/fn_chi2inv(): degrees of freedom df must be greater than 0.0");
+
+ if (p<=0.0) return (0.0);
+ else if (p>=1.0) return (MACHINEINFINITY);
+ else
+ return gsl_cdf_chisq_Pinv(p,df);
#else
***No default routine yet;
#endif
+}
//<<---------------
// Evaluating the standard normal cumulative distribution function.
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double fn_normalcdf(double x)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
return (imsls_d_normal_cdf(x));
-}
+#elif defined( USE_GSL_LIBRARY )
+ return gsl_cdf_ugaussian_P(x);
#else
***No default routine yet;
#endif
+}
+
//<<---------------
// Evaluating the inverse of the standard normal cumulative distribution function.
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double fn_normalinv(double p)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
return (imsls_d_normal_inverse_cdf(p));
-}
+#elif defined( USE_GSL_LIBRARY )
+ return gsl_cdf_ugaussian_Pinv(p);
#else
***No default routine yet;
#endif
+}
//<<---------------
// Evaluating the inverse of the beta cumulative distribution function.
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double fn_betainv(double p, double _alpha, double _beta)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
//p = int_{0}^{\infty} betapdf(t, _alpha, _beta) dt where betapdf(t,_alpha,_beta) \propt t^{_alpha-1}*(1-t)^(_beta-1}.
return (imsls_d_beta_inverse_cdf(p, _alpha, _beta));
-}
+#elif defined( USE_GSL_LIBRARY)
+ return gsl_cdf_beta_Pinv(p,_alpha,_beta);
#else
***No default routine yet;
#endif
+}
//<<---------------
// Computes log gamma (x) where gamma(n+1) = n! and gamma(x) = int_0^{\infty} e^{-t} t^{x-1} dt.
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double fn_gammalog(double x)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
return (imsl_d_log_gamma(x));
-}
+#elif defined( USE_GSL_LIBRARY )
+ return gsl_sf_lngamma(x);
#else
-No default routine yet;
+ ***No default routine yet;
#endif
+}
//<<---------------
// Computes log beta(x, y) where beta(x, y) = gamma(x)*gamm(y)/gamma(x+y).
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double fn_betalog(double x, double y)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
return (imsl_d_log_beta(x, y));
-}
+#elif defined( USE_GSL_LIBRARY )
+ return gsl_sf_lnbeta(x,y);
#else
***No default routine yet;
#endif
+}
//<<---------------
// Computes log gamma (x) where gamma(n+1) = n! and gamma(x) = int_0^{\infty} e^{-t} t^{x-1} dt.
//--------------->>
-#if defined( IMSL_STATISTICSTOOLBOX )
double gammalog(double x)
{
+#if defined( IMSL_STATISTICSTOOLBOX )
return (imsl_d_log_gamma(x));
-}
+#elif defined( USE_GSL_LIBRARY )
+ return gsl_sf_lngamma(x);
#else
-**No default routine yet;
+ ***No default routine yet;
#endif
+}
//-----------------------------------------------------------------------------------
diff --git a/CFiles/cstz.h b/CFiles/cstz.h
old mode 100755
new mode 100644
index e52a2bfe2c194269e888f9be33108d7574d6e8d0..c99c29ede23a04b3e343c113e0663566582c17b3
--- a/CFiles/cstz.h
+++ b/CFiles/cstz.h
@@ -1,7 +1,7 @@
#ifndef __CSTZ_H__
#define __CSTZ_H__
#include "tzmatlab.h"
- #include "switch_opt.h" //DW's Markov-switching routines, only used by gradcd_timet() and ComputeCovarianceFromOuterProduct().
+ #include "dw_switch_opt.h" //DW's Markov-switching routines, only used by gradcd_timet() and ComputeCovarianceFromOuterProduct().
typedef struct {
@@ -67,8 +67,6 @@
void gradfd_gen(double *g, double *x, int n, double (*fcn)(double *x, int n), double *grdh, double f0);
//=== For computing inverse Hessian.
- TSdmatrix *ComputeHessianFrom2ndDerivative(TSdmatrix *Hessian_dm, struct TStateModel_tag *smodel_ps, TSdvector *xhat_dv);
- //+
void gradcd_timet(TSdvector *g_dv, TSdvector *x_dv, int t, struct TStateModel_tag *smodel_ps, double (*fcn)(double *x, int t, struct TStateModel_tag *smodel_ps), double grdh, double f0);
TSdmatrix *ComputeHessianFromOuterProduct(TSdmatrix *Hessian_dm, struct TStateModel_tag *smodel_ps, TSdvector *xhat_dv);
TSdmatrix *ComputeCovarianceFromOuterProduct(TSdmatrix *Omega_dm, struct TStateModel_tag *smodel_ps, TSdvector *xhat_dv);
diff --git a/CFiles/cstz_dw.c b/CFiles/cstz_dw.c
old mode 100755
new mode 100644
diff --git a/CFiles/fn_filesetup.c b/CFiles/fn_filesetup.c
old mode 100755
new mode 100644
index f756f5c6c9ee394d86d4cc98a97187eef5ad955a..f75c865663be4beee65409b85ed646116a7fb90f
--- a/CFiles/fn_filesetup.c
+++ b/CFiles/fn_filesetup.c
@@ -14,7 +14,7 @@
int fn_ParseCommandLine(int n_arg, char **args, char ch) {
int i;
for (i=1; i<n_arg; i++)
- if ( ((args[i][0] == '/') || (args[i][0] == '-')) && (args[i][1] == ch) && (args[i][2] == '\0')) return i;
+ if ((args[i][0] == '/') && (args[i][1] == ch)) return i;
return 0;
}
@@ -30,12 +30,10 @@ int fn_ParseCommandLine(int n_arg, char **args, char ch) {
char *fn_ParseCommandLine_String(int n_arg, char **args, char ch, char *default_return) {
int i=fn_ParseCommandLine(n_arg,args,ch);
if (i > 0)
- //if (strlen(args[i]) > 2) return args[i]+2;
- // // In case the user forgot typing a space between /ch and string following it, still returns a pointer to the string folloing /ch.
- //else if ((i+1 < n_arg) && (args[i+1][0] != '/')) return args[i+1];
- // // Returns a pointer to the string that does NOT begin with / and there is a whitespace between /ch and the string.
- if (i+1 < n_arg) return args[i+1];
- // Returns a pointer to the string that does NOT begin with / and there is a whitespace between /ch and the string.
+ if (strlen(args[i]) > 2) return args[i]+2;
+ // In case the user forgot typing a space between /ch and string following it, still returns a pointer to the string folloing /ch.
+ else if ((i+1 < n_arg) && (args[i+1][0] != '/')) return args[i+1];
+ // Returns a pointer to the string that does NOT begin with / and there is a whitespace between /ch and the string.
return default_return;
}
diff --git a/CFiles/fn_filesetup.h b/CFiles/fn_filesetup.h
old mode 100755
new mode 100644
diff --git a/CFiles/gensys.c b/CFiles/gensys.c
old mode 100755
new mode 100644
diff --git a/CFiles/gensys.h b/CFiles/gensys.h
old mode 100755
new mode 100644
index ea1e2ac2f1ea3e3121606522becac2ba2588819b..9b6d3516d0522f1af9280517a4687130055d32ad
--- a/CFiles/gensys.h
+++ b/CFiles/gensys.h
@@ -48,7 +48,7 @@
TSdzmatrix *Gev_dzm; //n-by-2 z matrix of possible complex numbers.
TSivector *eu_iv; //2-by-1.
//=== Function itself.
- void (*gensys)(struct TSgensys_tag *, void *);
+ int (*gensys)(struct TSgensys_tag *, void *);
//=== Input arguments, which are all intialized to 0.0 and whose flags are set to M_GE.
TSdmatrix *G0_dm; //n-by-n.
TSdmatrix *G1_dm; //n-by-n.
diff --git a/CFiles/gensys01Mar06Works.c b/CFiles/gensys01Mar06Works.c
deleted file mode 100755
index d185626bf99db49e7bdbe4212f4704f7d1464c03..0000000000000000000000000000000000000000
--- a/CFiles/gensys01Mar06Works.c
+++ /dev/null
@@ -1,1114 +0,0 @@
-/*******************************************************************
- * [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
- *
- * System given as
- * g0*y(t)=g1*y(t-1)+c+psi*z(t)+pi*eta(t),
- * with z an exogenous variable process and eta being endogenously determined
- * one-step-ahead expectational errors. Returned system is
- * y(t)=G1*y(t-1)+C+impact*z(t)+ywt*inv(I-fmat*inv(L))*fwt*z(t+1) .
- * If z(t) is i.i.d., the last term drops out.
- * If div or stake is omitted from argument list, a div>1 or stake>1 is calculated.
- * eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
- * existence only with not-serially correlated z(t); eu=[-2,-2] for coincident zeros.
- *
- * g0, g1: n-by-n matrices.
- * c: n-by-1 constant terms.
- * z(t): m-by-1 vector of exogenous residuals where m < n.
- * psi: n-by-m matrix.
- * eta(t): h-by-1 vector of expectational (endogenous) errors.
- * pi: n-by-h matrix.
- * div: a real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- *-------
- * G1 or Theta_dm: n-by-n matrices.
- * C: n-by-1 vector of constant terms.
- * impact: n-by-m matrix.
- * gev: n-by-2 z vector of stacked generalized eigenvalues where gev(;,2) ./ gev(:,1) = eig(g0, g1).
- * ywt: n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- * fmat: nunstab-by-nunstab z matrix where nunstab is the number of non-stable roots.
- * fwt: nunstab-by-m z matrix.
- *
- * 1996 MATLAB algorithm by Christopher Sims
- * 2002 Mex implementation by Iskander Karibzhanov
- * 2004 Modified to C function by Tao Zha (April)
- *
- * Note: Iskander is transforming g0 and g1 to complex matrices and uses zgges() as a qz decomposition.
- * This is really wasting efficiency. One should keep g0 and g1 as real matrices and use
- * dgges() as a qz decomposition. I don't have time to overhaul this at this point. 04/20/04, T. Zha.
- * Note: 02/22/06. I take the above note back. According to DW, it is easy to order the
- * the generalized eigenvalues by using the complex g0 and g1. In principle, one could
- * order the roots using the real qz decomposition on real matrices g0 and g1. But so far
- * Dan has found it a pain to do it. Perhaps we should read the MKL Lapack manual more
- * carefully at a later point.
-********************************************************************/
-
-#include "gensys.h"
-
-
-//----- NOTE: We can't replace MKL_Complex16 with a different name because the Intel Lapack uses MKL_Complex16.
-//----- The only way to do this is to overhaul the code and put a wrapper function on each Intel Lapack function.
-static int selctg(MKL_Complex16 *alpha, MKL_Complex16 *beta);
-static int qz(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *q, MKL_Complex16 *z, int n);
-static MKL_Complex16* CreateComplexMatrix5RealMatrix(TSdmatrix *X_dm);
-static MKL_Complex16* CreateComplexMatrix5RealVector(TSdvector *x_dv);
-static void ComplexMatrix2RealMatrix(TSdmatrix *Y_dm, MKL_Complex16 *Z);
-static void ComplexMatrix2RealVector(TSdvector *y_dv, MKL_Complex16 *Z);
-static TSdzmatrix *SubComplexMatrix2Zmatrix(TSdzmatrix *X_dzm, MKL_Complex16 *Z, const int nrowsforZ, const int _m, const int _n);
-static void copy_eigenvalues(TSdzmatrix *Gev_dzm, MKL_Complex16 *a, MKL_Complex16 *b);
-static int compute_svd(MKL_Complex16 *a, MKL_Complex16 **u, double **d, MKL_Complex16 **v, int m, int n);
-static int compute_norm(MKL_Complex16 *a, double **d, int m, int n);
-//
-static int compute_normx(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *zwt, MKL_Complex16 *ueta, double **normx, int nunstab, int psin, int n, int bigev);
-static void cblas_zdupe(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdscali(int n, double *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdscale(int n, double *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdpsb(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb, MKL_Complex16 *c, int ldc);
-//
-static void InitializeConstantMLK_Complex16(MKL_Complex16 *x_clx, const int _n, const double c);
-static void InitializeConstantDouble(double *x_p, const int _n, const double c);
-
-
-
-TSgensys *CreateTSgensys(TFlinratexp *func, const int _n, const int _m, const int _k, const double div)
-{
- //_n is the number of stacked variables (endogenous, Lagurangian multiplier, expected multiplier, etc.).
- //_m is the number of exogenous shocks.
- //_k is the number of expectational errors.
- //div is the dividing number to determine what constitutes an unstable root. If div<1.0, a div>1.0 is calculated mechanically.
- TSgensys *gensys_ps = tzMalloc(1, TSgensys);
-
- //=== Output arguments.
- gensys_ps->Theta_dm = CreateMatrix_lf(_n, _n); //n-by-n.
- gensys_ps->c_dv = CreateVector_lf(_n); //n-by-1.
- gensys_ps->Impact_dm = CreateMatrix_lf(_n, _m); //n-by-m.
- gensys_ps->Fmat_dzm = (TSdzmatrix *)NULL; //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- gensys_ps->Fwt_dzm = (TSdzmatrix *)NULL; //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- gensys_ps->Ywt_dzm = (TSdzmatrix *)NULL; //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- gensys_ps->Gev_dzm = CreateMatrix_dz(_n, 2); //n-by-2 z matrix of possible complex numbers.
- gensys_ps->eu_iv = CreateConstantVector_int(2, 0); //2-by-1.
-
- //=== Function itself.
- gensys_ps->gensys = func;
-
- //=== Input arguments.
- gensys_ps->G0_dm = CreateConstantMatrix_lf(_n, _n, 0.0); //n-by-n.
- gensys_ps->G0_dm->flag = M_GE;
- gensys_ps->G1_dm = CreateConstantMatrix_lf(_n, _n, 0.0); //n-by-n.
- gensys_ps->G1_dm->flag = M_GE;
- gensys_ps->c0_dv = CreateConstantVector_lf(_n, 0.0); //n-by-1.
- gensys_ps->Psi_dm = CreateConstantMatrix_lf(_n, _m, 0.0); //n-by-m.
- gensys_ps->Psi_dm->flag = M_GE;
- gensys_ps->Pi_dm = CreateConstantMatrix_lf(_n, _k, 0.0); //n-by-k where k is the number of expectational errors.
- gensys_ps->Pi_dm->flag = M_GE;
- gensys_ps->div = div;
-
- return (gensys_ps);
-}
-//-------
-TSgensys *DestroyTSgensys(TSgensys *gensys_ps)
-{
- if (gensys_ps) {
- //=== Output arguments.
- DestroyMatrix_lf(gensys_ps->Theta_dm); //n-by-n.
- DestroyVector_lf(gensys_ps->c_dv); //n-by-1.
- DestroyMatrix_lf(gensys_ps->Impact_dm); //n-by-m.
- DestroyMatrix_dz(gensys_ps->Fmat_dzm); //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- DestroyMatrix_dz(gensys_ps->Fwt_dzm); //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- DestroyMatrix_dz(gensys_ps->Ywt_dzm); //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- DestroyMatrix_dz(gensys_ps->Gev_dzm); //n-by-2 z matrix of possible complex numbers.
- DestroyVector_int(gensys_ps->eu_iv); //2-by-1.
-
- //=== Input arguments.
- DestroyMatrix_lf(gensys_ps->G0_dm); //n-by-n.
- DestroyMatrix_lf(gensys_ps->G1_dm); //n-by-n.
- DestroyVector_lf(gensys_ps->c0_dv); //n-by-1.
- DestroyMatrix_lf(gensys_ps->Psi_dm); //n-by-m.
- DestroyMatrix_lf(gensys_ps->Pi_dm); //n-by-k where k is the number of expectational errors.
-
- free(gensys_ps);
-
- return ((TSgensys *)NULL);
- }
- else return (gensys_ps);
-}
-
-
-//--------------------------- For the function gensys_sims() ------------------------------------
-static int fixdiv = 1, zxz = 0;
-static double stake = 1.01;
-static int nunstab = 0;
-static MKL_Complex16 one, minusone, zero;
-
-/* [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensysmkl(g0,g1,c,psi,pi,stake) */
-//void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
-void gensys_sims(TSgensys *gensys_ps, void *dummy_ps)
-{
- int tmpi;
- int n, psin, pin, nsquare, md, mds, md1, i, bigev, bigevs, bigev1;
- int *eu;
- int exist = 0, existx = 0, unique = 0;
- //=== Memory will be allocated to the following.
- double *deta = NULL, *dz = NULL, *normx = NULL, *deta1 = NULL, *norm = NULL;
- MKL_Complex16 *a = NULL, *b = NULL, *q = NULL, *z = NULL, *pi = NULL, *psi = NULL;
- MKL_Complex16 *tmat = NULL, *g0 = NULL, *g1 = NULL, *dummy = NULL, *tmatq = NULL, *c = NULL, *impact = NULL, *ab = NULL;
- MKL_Complex16 *fmat = NULL, *fwt = NULL, *ywt = NULL;
- MKL_Complex16 *etawt = NULL, *ueta = NULL, *veta = NULL, *zwt = NULL, *uz = NULL, *vz = NULL, *etawt1 = NULL, *ueta1 = NULL, *veta1 = NULL;
-
-// /* Check for proper number of input and output arguments */
-// if (nrhs < 5 || nrhs > 6)
-// mexErrMsgTxt("Five or six input arguments required.\n");
-// if (nlhs != 8)
-// mexErrMsgTxt("Eight output arguments required.\n");
-
-// /* Dimensions */
-// n = mxGetM(prhs[0]);
-// psin = mxGetN(prhs[3]);
-// pin = mxGetN(prhs[4]);
- n = gensys_ps->G0_dm->nrows;
- psin = gensys_ps->Psi_dm->ncols;
- pin = gensys_ps->Pi_dm->ncols;
- //
- eu = gensys_ps->eu_iv->v;
- //eu = mxGetPr(plhs[7]=mxCreateDoubleMatrix(2,1,mxREAL));
-
-// /* Check for consistency of input matrix dimensions */
-// if (mxGetM(prhs[0])!=n || mxGetM(prhs[1])!=n)
-// mexErrMsgTxt("g0 and g1 must be square matrices.\n");
-
-// if (mxGetM(prhs[3])!=n || mxGetM(prhs[4])!=n)
-// mexErrMsgTxt("psi and pi must have the same number of rows as g0.\n");
-
-// eu = mxGetPr(plhs[7]=mxCreateDoubleMatrix(2,1,mxREAL));
-
- //=== [a b q z]=qz(g0,g1);
-// a = mat2mkl(prhs[0],n,n);
-// b = mat2mkl(prhs[1],n,n);
-// q = mxCalloc(n*n,sizeof(MKL_Complex16));
-// z = mxCalloc(n*n,sizeof(MKL_Complex16));
- a = CreateComplexMatrix5RealMatrix(gensys_ps->G0_dm);
- b = CreateComplexMatrix5RealMatrix(gensys_ps->G1_dm);
- q = tzMalloc(nsquare=square(n), MKL_Complex16);
- z = tzMalloc(nsquare, MKL_Complex16);
- InitializeConstantMLK_Complex16(q, nsquare, 0.0);
- InitializeConstantMLK_Complex16(z, nsquare, 0.0);
-
-// fixdiv = nrhs != 6;
-// stake = fixdiv ? 1.01 : mxGetScalar(prhs[5]);
- fixdiv = (gensys_ps->div < 1.0);
- stake = fixdiv ? 1.01 : gensys_ps->div;
- nunstab = 0;
- zxz = 0;
-
- if (qz(a, b, q, z, n)) {
- printf("WARNING: QZ factorization failed.\n");
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
-
- nunstab /= 2;
-
- if (zxz) {
- printf("WARNING: Coincident zeros. Indeterminacy and/or nonexistence.\n");
- eu[0] = eu[1] = -2;
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
-
-// plhs[6] = mxCreateDoubleMatrix(n, 2, mxCOMPLEX); //For Gev_dzm.
-// copy_eigenvalues(plhs[6], a, b, n);
- copy_eigenvalues(gensys_ps->Gev_dzm, a, b);
-
- one.real = 1.0;
- one.imag = 0.0;
-
- minusone.real = -1.0;
- minusone.imag = 0.0;
-
- zero.real = 0.0;
- zero.imag = 0.0;
-
-// pi = mat2mkl(prhs[4],n,pin);
-// etawt = mxCalloc(nunstab*pin,sizeof(MKL_Complex16));
- pi = CreateComplexMatrix5RealMatrix(gensys_ps->Pi_dm);
- etawt = tzMalloc(tmpi=nunstab*pin, MKL_Complex16);
- InitializeConstantMLK_Complex16(etawt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, pin, n,
- &one, q+n*(n-nunstab), n, pi, n, &zero, etawt, nunstab);
- if (compute_svd(etawt, &ueta, &deta, &veta, nunstab, pin)) {
- //Memory is now allocated to ueta, deta, and veta.
- printf("WARNING: SVD failed.\n");
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(etawt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(etawt);
- md = nunstab<pin?nunstab:pin;
- bigev = md;
- for (i=0; i<md; i++)
- if (deta[i]<=REALSMALL) {
- bigev=i;
- break;
- }
-
-// psi = mat2mkl(prhs[3],n,psin);
-// zwt = mxCalloc(nunstab*psin,sizeof(MKL_Complex16));
- psi = CreateComplexMatrix5RealMatrix(gensys_ps->Psi_dm);
- zwt = tzMalloc(tmpi=nunstab*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(zwt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, psin, n,
- &one, q+n*(n-nunstab), n, psi, n, &zero, zwt, nunstab);
- if (compute_svd(zwt, &uz, &dz, &vz, nunstab, psin)) {
- //Memory is now allocated to uz, dz, and vz.
- printf("WARNING: SV decomposition failed.\n");
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(uz);
- tzDestroy(dz);
- tzDestroy(vz);
- tzDestroy(zwt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(vz);
- mds = nunstab<psin?nunstab:psin;
- bigevs = mds;
- for (i=0; i<mds; i++)
- if (dz[i]<=REALSMALL) {
- bigevs=i;
- break;
- }
- tzDestroy(dz);
-
-
- /* ueta = nunstab x bigev
- deta = bigev x 1
- veta = bigev x pin, ldveta = md
- uz = nunstab x bigevs
- dz = bigevs x 1
- vz = bigevs x psin, ldvz = mds */
-
- if (!bigevs) {
- exist = 1;
- existx = 1;
- } else {
- /* uz-ueta*ueta'*uz */
- MKL_Complex16 *tmp = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, nunstab, nunstab,
- bigev, &one, ueta, nunstab, ueta, nunstab, &zero, tmp, nunstab);
- cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, nunstab,
- bigevs, &minusone, tmp, nunstab, uz, nunstab, &one, uz, nunstab);
- tzDestroy(tmp);
- if (compute_norm(uz, &norm, nunstab, bigevs)) {
- //Memory is now allocated to norm.
- printf("WARNING: SVD failed.\n");
- tzDestroy(norm);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(uz);
- tzDestroy(zwt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- exist = *norm < REALSMALL*n;
- tzDestroy(norm);
- if (compute_normx(a, b, zwt, ueta, &normx, nunstab, psin, n, bigev)) {
- //If 0, memory is now allocated to normx; otherwise, normx is destroyed within the function compute_normx().
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(uz);
- tzDestroy(zwt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- existx = *normx < REALSMALL*n;
- tzDestroy(normx);
- }
-
- tzDestroy(uz);
- tzDestroy(zwt);
-
-
-/* Note that existence and uniqueness are not just matters of comparing
- numbers of roots and numbers of endogenous errors. These counts are
- reported below because usually they point to the source of the problem. */
-
- etawt1 = tzMalloc(tmpi=(n-nunstab)*pin, MKL_Complex16);
- InitializeConstantMLK_Complex16(etawt1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, n-nunstab, pin, n,
- &one, q, n, pi, n, &zero, etawt1, n-nunstab);
- tzDestroy(pi);
- if (compute_svd(etawt1, &ueta1, &deta1, &veta1, n-nunstab, pin)) {
- //Memory is now allocated to ueta1, deta1, and veta1.
- printf("WARNING: SVD failed for compute_svd().\n");
- tzDestroy(ueta1);
- tzDestroy(deta1);
- tzDestroy(veta1);
- tzDestroy(etawt1);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(etawt1);
- md1 = n-nunstab<pin?n-nunstab:pin;
- bigev1 = md1;
- for (i=0; i<md1; i++)
- if (deta1[i]<=REALSMALL) {
- bigev1=i;
- break;
- }
- /* ueta1 = n-nunstab x bigev1
- deta1 = bigev1 x 1
- veta1 = bigev1 x pin, ldveta1 = md1 */
-
- if (existx || !nunstab) {
- //=== Solution exists.
- eu[0] = 1;
- } else {
- if (exist) {
- printf("WARNING: Solution exists for unforecastable z only\n");
- eu[0] = -1;
- } /* else
- mexPrintf("No solution. %d unstable roots. %d endog errors.\n",nunstab,bigev1); */
- /* mexPrintf("Generalized eigenvalues\n");
- mexCallMATLAB(0,NULL,1, &plhs[6], "disp"); */
- }
- if (!bigev1)
- unique = 1;
- else {
- /* veta1-veta1*veta*veta' */
- /* veta = bigev x pin, ldveta1 = md
- veta1 = bigev1 x pin, ldveta1 = md1 */
- MKL_Complex16 *tmp = tzMalloc(pin*pin, MKL_Complex16);
- MKL_Complex16 *veta1_copy = tzMalloc(pin*bigev1, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, pin*pin, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- InitializeConstantMLK_Complex16(veta1_copy, pin*bigev1, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, pin, pin,
- bigev, &one, veta, md, veta, md, &zero, tmp, pin);
- cblas_zdupe(bigev1,pin,veta1,md1,veta1_copy,bigev1);
- cblas_zhemm(CblasColMajor, CblasRight, CblasUpper, bigev1, pin,
- &minusone, tmp, pin, veta1_copy, bigev1, &one, veta1_copy, bigev1);
- tzDestroy(tmp);
- if (compute_norm(veta1_copy, &norm, bigev1, pin)) {
- //Memory is now allocated to norm.
- printf("WARNING: SVD failed.\n");
- tzDestroy(norm);
- tzDestroy(ueta1);
- tzDestroy(deta1);
- tzDestroy(veta1);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(veta1_copy);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(veta1_copy);
- unique = *norm < REALSMALL*n;
- tzDestroy(norm);
- }
- if (unique) {
- //=== Unique solution.
- eu[1] = 1;
- } else {
- eu[1] = 0;
- printf("WARNING: Indeterminacy. %d loose endog errors with eu being [%d, %d].\n",bigev1-bigev, eu[0], eu[1]);
- //printf("WARNING: Indeterminacy. %d loose endog errors with eu being [%g, %g].\n",bigev1-bigev, gensys_ps->eu_dv->v[0], gensys_ps->eu_dv->v[1]);
- //printf("WARNING: Indeterminacy. %d loose endog errors.\n",bigev1-bigev);
- /* mexPrintf("Generalized eigenvalues\n");
- mexCallMATLAB(0,NULL,1, &plhs[6], "disp"); */
- }
-
- cblas_zdscali(pin,deta,bigev,veta,md); /* veta' = deta\veta' */
- tzDestroy(deta);
- cblas_zdscale(pin,deta1,bigev1,veta1,md1); /* veta1' = deta1*veta1' */
- tzDestroy(deta1);
- etawt = tzMalloc(tmpi=nunstab*pin, MKL_Complex16); /* etawt = ueta*veta' */
- InitializeConstantMLK_Complex16(etawt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, nunstab, pin, bigev,
- &one, ueta, nunstab, veta, md, &zero, etawt, nunstab);
- tzDestroy(ueta);
- tzDestroy(veta);
- etawt1 = tzMalloc(tmpi=(n-nunstab)*pin, MKL_Complex16); /* etawt1 = ueta1*veta1' */
- InitializeConstantMLK_Complex16(etawt1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, pin, bigev1,
- &one, ueta1, n-nunstab, veta1, md1, &zero, etawt1, n-nunstab);
- tzDestroy(ueta1);
- tzDestroy(veta1);
- tmat = tzMalloc(tmpi=(n-nunstab)*nunstab, MKL_Complex16); /* tmat = etawt1*etawt' */
- InitializeConstantMLK_Complex16(tmat, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n-nunstab, nunstab, pin,
- &one, etawt1, n-nunstab, etawt, nunstab, &zero, tmat, n-nunstab);
- tzDestroy(etawt1);
- tzDestroy(etawt);
-
- g0 = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(g0, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdupe(n-nunstab, n, a, n, g0, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, nunstab, nunstab,
- &minusone, tmat, n-nunstab, a+(n-nunstab)*(n+1), n, &one, g0+(n-nunstab)*n, n);
- cblas_zcopy(nunstab, &one, 0, g0+(n-nunstab)*(n+1), n+1);
-
- g1 = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(g1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdupe(n-nunstab, n, b, n, g1, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, nunstab, nunstab,
- &minusone, tmat, n-nunstab, b+(n-nunstab)*(n+1), n, &one, g1+(n-nunstab)*n, n);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, n, &one, g0, n, g1, n);
- dummy = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, n, n, &one, z, n, g1, n, &zero, dummy, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n, n, n, &one, dummy, n, z, n, &zero, g1, n);
- tzDestroy(dummy);
- //plhs[0] = mkl2mat_real(g1, n, n, n);
- ComplexMatrix2RealMatrix(gensys_ps->Theta_dm, g1); //Output.
- tzDestroy(g1);
-
- tmatq = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmatq, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zcopy(n*n, q, 1, tmatq, 1);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n, n-nunstab, nunstab,
- &minusone, tmatq+(n-nunstab)*n, n, tmat, n-nunstab, &one, tmatq, n);
- tzDestroy(tmat);
-
- ab = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(ab, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdpsb(nunstab, nunstab, a+(n-nunstab)*(n+1), n, b+(n-nunstab)*(n+1), n, ab, nunstab);
- cblas_ztrsm(CblasColMajor, CblasRight, CblasUpper, CblasConjTrans, CblasNonUnit,
- n, nunstab, &one, ab, nunstab, tmatq+(n-nunstab)*n, n);
- tzDestroy(ab);
-
- //c = mat2mkl(prhs[2],n,1);
- c = CreateComplexMatrix5RealVector(gensys_ps->c0_dv);
- dummy = tzMalloc(gensys_ps->c0_dv->n, MKL_Complex16);
- //$$$$$$$ The following is Iskander's fatal code error. One cannot use c in the two different places; otherwise, it makes c be zero completely!
- // cblas_zgemv(CblasColMajor, CblasConjTrans, n, n, &one, tmatq, n, c, 1, &zero, c, 1);
- // cblas_ztrsv(CblasColMajor, CblasUpper, CblasNoTrans, CblasNonUnit, n, g0, n, c, 1);
- // cblas_zgemv(CblasColMajor, CblasNoTrans, n, n, &one, z, n, c, 1, &zero, c, 1);
- cblas_zgemv(CblasColMajor, CblasConjTrans, n, n, &one, tmatq, n, c, 1, &zero, dummy, 1);
- cblas_ztrsv(CblasColMajor, CblasUpper, CblasNoTrans, CblasNonUnit, n, g0, n, dummy, 1);
- cblas_zgemv(CblasColMajor, CblasNoTrans, n, n, &one, z, n, dummy, 1, &zero, c, 1);
- //plhs[1] = mkl2mat_real(c, n, n, 1);
- ComplexMatrix2RealVector(gensys_ps->c_dv, c); //Output.
- tzDestroy(c);
- tzDestroy(dummy);
-
- impact = tzMalloc(tmpi=n*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(impact, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, n-nunstab, psin, n,
- &one, tmatq, n, psi, n, &zero, impact, n);
- tzDestroy(tmatq);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, psin, &one, g0, n, impact, n);
- dummy = tzMalloc(tmpi=n*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, psin, n,
- &one, z, n, impact, n, &zero, dummy, n);
- tzDestroy(impact);
- //plhs[2] = mkl2mat_real(dummy, n, n, psin);
- ComplexMatrix2RealMatrix(gensys_ps->Impact_dm, dummy); //Output.
- tzDestroy(dummy);
-
- fmat = a+(n-nunstab)*(n+1);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, nunstab, &one, b+(n-nunstab)*(n+1), n, fmat, n);
- //plhs[3] = mkl2mat(fmat, n, nunstab, nunstab);
- gensys_ps->Fmat_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Fmat_dzm, fmat, n, nunstab, nunstab);
- tzDestroy(a);
-
- fwt = tzMalloc(tmpi=nunstab*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(fwt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_ztrsm(CblasColMajor, CblasRight, CblasUpper, CblasConjTrans, CblasNonUnit,
- n, nunstab, &one, b+(n-nunstab)*(n+1), n, q+(n-nunstab)*n, n);
- tzDestroy(b);
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, psin, n,
- &minusone, q+(n-nunstab)*n, n, psi, n, &zero, fwt, nunstab);
- tzDestroy(q);
- tzDestroy(psi);
- //plhs[4] = mkl2mat(fwt, nunstab, nunstab, psin);
- gensys_ps->Fwt_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Fwt_dzm, fwt, nunstab, nunstab, psin);
- tzDestroy(fwt);
-
- ywt = tzMalloc(tmpi=n*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(ywt, tmpi, 0.0);
- cblas_zcopy(nunstab, &one, 0, ywt+n-nunstab, n+1);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, nunstab, &one, g0, n, ywt, n);
- tzDestroy(g0);
- dummy = tzMalloc(tmpi=n*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, nunstab, n,
- &one, z, n, ywt, n, &zero, dummy, n);
- tzDestroy(z);
- tzDestroy(ywt);
- //plhs[5] = mkl2mat(dummy, n, n, nunstab);
- gensys_ps->Ywt_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Ywt_dzm, dummy, n, n, nunstab);
- tzDestroy(dummy);
-}
-
-
-/******************************************************************************
- * function selctg orders the eigenvalues so that a selected cluster of *
- * eigenvalues appears in the leading diagonal blocks of the upper *
- * quasi-triangular matrix S and the upper triangular matrix T. *
- ******************************************************************************/
-
-static int selctg(MKL_Complex16 *alpha, MKL_Complex16 *beta)
-{
- double absA = sqrt(alpha->real*alpha->real+alpha->imag*alpha->imag),
- absB = fabs(beta->real);
- if (absA) {
- double divhat = absB/absA;
- //bug detected by Vasco Curdia and Daria Finocchiaro, 2/25/2004 CAS. A root of
- //exactly 1.01 and no root between 1 and 1.02, led to div being stuck at 1.01
- //and the 1.01 root being misclassified as stable. Changing < to <= below fixes this.
- if (fixdiv && 1+REALSMALL<divhat && divhat<=stake)
- stake = (1+divhat)/2;
- }
- if (absA<REALSMALL && absB<REALSMALL)
- zxz = 1;
- if (absB>stake*absA) {
- nunstab++;
- return(0);
- } else
- return(1);
-}
-
-/******************************************************************************
- * compute for a pair of N-by-N complex nonsymmetric matrices (A,B), *
- * the generalized eigenvalues, the generalized complex Schur form (S, T), *
- * and optionally left and/or right Schur vectors (VSL and VSR) *
- ******************************************************************************/
-
-static int qz(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *q, MKL_Complex16 *z, int n)
-{
-// unsigned char msg[101];
- int sdim, lwork = -1, info = 0;
- MKL_Complex16 *alpha = tzMalloc(n,MKL_Complex16),
- *beta = tzMalloc(n,MKL_Complex16),
- *work, work1;
- double *rwork = tzMalloc(8*n, double);
- int *bwork = tzMalloc(4*n, int);
-
- /* Query zgges on the value of lwork */
- zgges("V", "V", "S", &selctg, &n, a, &n, b, &n, &sdim, alpha, beta, q,
- &n, z, &n, &work1, &lwork, rwork, bwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to the Intel MKL function zgges() has an illegal value",-info);
- tzDestroy(bwork);
- tzDestroy(rwork);
- tzDestroy(alpha);
- tzDestroy(beta);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgges("V", "V", "S", &selctg, &n, a, &n, b, &n, &sdim, alpha, beta, q,
- &n, z, &n, work, &lwork, rwork, bwork, &info);
-
- tzDestroy(work);
- tzDestroy(bwork);
- tzDestroy(rwork);
- tzDestroy(alpha);
- tzDestroy(beta);
-
- if (info < 0) {
- printf("WARNING: Input %d to the Intel MKL function zgges() has an illegal value",-info);
- return(info);
- }
-
- if (info > 0)
- if (info < n)
- printf("WARNING: The QZ iteration failed. (A,B) are not in Schur form,\n"
- "but ALPHA(j) and BETA(j) should be correct for j=%d,...,N.",info+1);
- else {
- switch (info-n) {
- case 1:
- printf("WARNING: LAPACK problem: error return from ZGGBAL");
- break;
- case 2:
- printf("WARNING: LAPACK problem: error return from ZGEQRF");
- break;
- case 3:
- printf("WARNING: LAPACK problem: error return from ZUNMQR");
- break;
- case 4:
- printf("WARNING: LAPACK problem: error return from ZUNGQR");
- break;
- case 5:
- printf("WARNING: LAPACK problem: error return from ZGGHRD");
- break;
- case 6:
- printf("WARNING: LAPACK problem: error return from ZHGEQZ (other than failed iteration)");
- break;
- case 7:
- printf("WARNING: LAPACK problem: error return from ZGGBAK (computing VSL)");
- break;
- case 8:
- printf("WARNING: LAPACK problem: error return from ZGGBAK (computing VSR)");
- break;
- case 9:
- printf("WARNING: LAPACK problem: error return from ZLASCL (various places)");
- break;
- default:
- printf("WARNING: LAPACK problem: unknown error.");
- break;
- }
- }
- return(info);
-}
-
-/*
- * Convert MATLAB complex matrix to MKL complex storage.
- *
- * Z = mat2mkl(X,ldz,ndz)
- *
- * converts MATLAB's mxArray X to MKL_Complex16 Z(ldz,ndz).
- * The parameters ldz and ndz determine the storage allocated for Z,
- * while mxGetM(X) and mxGetN(X) determine the amount of data copied.
- */
-
-//MKL_Complex16* mat2mkl(const mxArray *X, int ldz, int ndz) {
-// MKL_Complex16 *Z, *zp;
-// int m, n, incz, cmplxflag;
-// register int i, j;
-// double *xr, *xi;
-
-// Z = mxCalloc(ldz*ndz, sizeof(MKL_Complex16));
-// xr = mxGetPr(X);
-// xi = mxGetPi(X);
-// m = mxGetM(X);
-// n = mxGetN(X);
-// zp = Z;
-// incz = ldz-m;
-// cmplxflag = (xi != NULL);
-// for (j = 0; j < n; j++) {
-// if (cmplxflag) {
-// for (i = 0; i < m; i++) {
-// zp->real = *xr++;
-// zp->imag = *xi++;
-// zp++;
-// }
-// } else {
-// for (i = 0; i < m; i++) {
-// zp->real = *xr++;
-// zp++;
-// }
-// }
-// zp += incz;
-// }
-// return(Z);
-//}
-
-
-/*
- * Convert MKL complex storage to MATLAB real and imaginary parts.
- *
- * X = mkl2mat(Z,ldz,m,n)
- *
- * copies MKL_Complex16 Z to X, producing a complex mxArray
- * with mxGetM(X) = m and mxGetN(X) = n.
- */
-
-//mxArray* mkl2mat(MKL_Complex16 *Z, int ldz, int m, int n) {
-// int i, j, incz;
-// double *xr, *xi;
-// MKL_Complex16 *zp;
-// mxArray *X;
-
-// X = mxCreateDoubleMatrix(m,n,mxCOMPLEX);
-// xr = mxGetPr(X);
-// xi = mxGetPi(X);
-// zp = Z;
-// incz = ldz-m;
-// for (j = 0; j < n; j++) {
-// for (i = 0; i < m; i++) {
-// *xr++ = zp->real;
-// *xi++ = zp->imag;
-// zp++;
-// }
-// zp += incz;
-// }
-// return(X);
-//}
-
-//plhs[3] = mkl2mat(fmat, n, nunstab, nunstab)
-
-/*
- * Convert MKL complex storage to MATLAB real matrix ignoring imaginary part.
- *
- * X = mkl2mat(Z,ldz,m,n)
- *
- * copies MKL_Complex16 Z to X, producing a real mxArray
- * with mxGetM(X) = m and mxGetN(X) = n.
- */
-
-//mxArray* mkl2mat_real(MKL_Complex16 *Z, int ldz, int m, int n) {
-// int i, j, incz;
-// double *xr;
-// MKL_Complex16 *zp;
-// mxArray *X;
-
-// X = mxCreateDoubleMatrix(m,n,mxREAL);
-// xr = mxGetPr(X);
-// zp = Z;
-// incz = ldz-m;
-// for (j = 0; j < n; j++) {
-// for (i = 0; i < m; i++) {
-// *xr++ = zp->real;
-// zp++;
-// }
-// zp += incz;
-// }
-// return(X);
-//}
-
-//void copy_eigenvalues(mxArray *gev, MKL_Complex16 *a, MKL_Complex16 *b, int n) {
-// double *gevr = mxGetPr(gev),
-// *gevi = mxGetPi(gev);
-// int i;
-
-// for (i=0; i<n; i++, gevr++, gevi++, a+=n+1) {
-// *gevr = a->real;
-// *gevi = a->imag;
-// }
-
-// for (i=0; i<n; i++, gevr++, gevi++, b+=n+1) {
-// *gevr = b->real;
-// *gevi = b->imag;
-// }
-//}
-
-static void copy_eigenvalues(TSdzmatrix *Gev_dzm, MKL_Complex16 *a, MKL_Complex16 *b)
-{
- int n = Gev_dzm->real->nrows;
- double *gevr = Gev_dzm->real->M,
- *gevi = Gev_dzm->imag->M;
- int i;
-
- for (i=0; i<n; i++, gevr++, gevi++, a+=n+1) {
- *gevr = a->real;
- *gevi = a->imag;
- }
-
- for (i=0; i<n; i++, gevr++, gevi++, b+=n+1) {
- *gevr = b->real;
- *gevi = b->imag;
- }
-}
-
-
-static int compute_svd(MKL_Complex16 *x, MKL_Complex16 **u, double **d, MKL_Complex16 **v, int m, int n)
-{
- //$$$Memory allocated to u, d, and v will be destroyed outside this function.$$$
- int tmpi;
- int md = m<n?m:n, lwork = -1, info = 0;
- MKL_Complex16 *a, *work, work1;
- double *rwork = tzMalloc(5*md>1?5*md:1, double);
-
- a = tzMalloc(m*n, MKL_Complex16);
- cblas_zcopy(m*n, x, 1, a, 1);
-
- *u = tzMalloc(tmpi=m*md,MKL_Complex16);
- InitializeConstantMLK_Complex16(*u, tmpi, 0.0);
- *v = tzMalloc(tmpi=md*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(*v, tmpi, 0.0);
- *d = tzMalloc(md, double);
- InitializeConstantDouble(*d, md, 0.0);
-
- /* Query zgges on the value of lwork */
- zgesvd("S", "S", &m, &n, a, &m, *d, *u, &m, *v, &md, &work1, &lwork, rwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
- tzDestroy(rwork);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgesvd("S", "S", &m, &n, a, &m, *d, *u, &m, *v, &md, work, &lwork, rwork, &info);
-
- tzDestroy(work);
- tzDestroy(rwork);
- tzDestroy(a);
-
- if (info < 0)
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
-
- if (info > 0)
- printf("WARNING: ZBDSQR did not converge.\n%d superdiagonals of an intermediate "
- "bidiagonal form B did not converge to zero.",info);
- return(info);
-}
-
-static int compute_norm(MKL_Complex16 *a, double **d, int m, int n)
-{
- //Memory will be allocated to d, which will be destroyed outside this function.
- int md = m<n?m:n, lwork = -1, info = 0;
- MKL_Complex16 *work = NULL, work1;
- double *rwork = tzMalloc(5*md>1?5*md:1, double);
-
- *d = tzMalloc(md, double);
-
- /* Query zgges on the value of lwork */
- zgesvd("N", "N", &m, &n, a, &m, *d, NULL, &m, NULL, &md, &work1, &lwork, rwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
- tzDestroy(rwork);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgesvd("N", "N", &m, &n, a, &m, *d, NULL, &m, NULL, &md, work, &lwork, rwork, &info);
-
- tzDestroy(work);
- tzDestroy(rwork);
-
- if (info < 0)
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
-
- if (info > 0)
- printf("WARNING: ZBDSQR() in Intel MKL did not converge.\n%d superdiagonals of an intermediate "
- "bidiagonal form B did not converge to zero.",info);
-
- return(info);
-}
-
-static int compute_normx(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *zwt, MKL_Complex16 *ueta, double **normx, int nunstab, int psin, int n, int bigev)
-{
- //Memory is allocated to normx, which will be freed outside this function.
- int tmpi;
- int info = 0, i, bigevs;
- //
- MKL_Complex16 *M = NULL, *zwtx = NULL, *ux = NULL, *vx = NULL, *tmp = NULL;
- double *dx = NULL;
-
-
- a += (n+1)*(n-nunstab);
- b += (n+1)*(n-nunstab);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, psin, &one, b, n, zwt, nunstab);
- M = tzMalloc(nunstab*nunstab, MKL_Complex16);
- cblas_zdupe(nunstab, nunstab, a, n, M, nunstab);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, nunstab, &one, b, n, M, nunstab);
-
- zwtx = tzMalloc(nunstab*nunstab*psin, MKL_Complex16);
- cblas_zcopy(nunstab*psin, zwt, 1, zwtx, 1);
- for (i=1; i<nunstab; i++) {
- cblas_ztrmm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, nunstab, psin*i, &one, M, nunstab, zwtx, nunstab);
- cblas_zcopy(nunstab*psin, zwt, 1, zwtx+nunstab*psin*i, 1);
- }
- tzDestroy(M);
- cblas_ztrmm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, nunstab, nunstab*psin, &one, b, n, zwtx, nunstab);
- info = compute_svd(zwtx, &ux, &dx, &vx, nunstab, nunstab*psin); //Memory is allocated to ux, dx, and vx.
- tzDestroy(vx);
- tzDestroy(zwtx);
- if (info) {
- printf("WARNING: SVD failed.\n");
- tzDestroy(ux);
- tzDestroy(dx);
- return(info);
- }
- bigevs = nunstab;
- for (i=0; i<nunstab; i++)
- if (dx[i]<=REALSMALL) {
- bigevs = i;
- break;
- }
- tzDestroy(dx);
- /* ux-ueta*ueta'*ux */
- tmp = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, nunstab, nunstab,
- bigev, &one, ueta, nunstab, ueta, nunstab, &zero, tmp, nunstab);
- cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, nunstab,
- bigevs, &minusone, tmp, nunstab, ux, nunstab, &one, ux, nunstab);
- tzDestroy(tmp);
- info = compute_norm(ux, normx, nunstab, bigevs); //Memory is allocated to normx.
- if (info) {
- printf("WARNING: SVD failed.\n");
- tzDestroy(normx);
- tzDestroy(ux);
- return(info);
- }
- tzDestroy(ux);
- return(info);
-}
-
-static void cblas_zdupe(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<m; i++, a++, b++)
- cblas_zcopy(n, a, lda, b, ldb);
-}
-
-static void cblas_zdscali(int n, double *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<lda; i++, a++, b++)
- cblas_zdscal(n, 1.0/(*a), b, ldb);
-}
-
-static void cblas_zdscale(int n, double *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<lda; i++, a++, b++)
- cblas_zdscal(n, *a, b, ldb);
-}
-
-static void cblas_zdpsb(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb, MKL_Complex16 *c, int ldc)
-{
- int i;
- cblas_zdupe(m, n, a, lda, c, ldc);
- for (i=0; i<m; i++, b++, c++)
- cblas_zaxpy(n, &minusone, b, ldb, c, ldc);
-}
-
-
-static MKL_Complex16* CreateComplexMatrix5RealMatrix(TSdmatrix *X_dm)
-{
- int mn, k;
- double *M;
- //
- MKL_Complex16 *Z = NULL;
-
- if (!X_dm) fn_DisplayError("CreateComplexMatrix5RealMatrix(): Input matrix X_dm must be allocated memory");
- M = X_dm->M;
-
- Z = tzMalloc(mn=X_dm->nrows*X_dm->ncols, MKL_Complex16);
- for (k=mn-1; k>=0; k--) {
- Z[k].real = M[k];
- Z[k].imag = 0.0;
- }
- return(Z);
-}
-
-static MKL_Complex16* CreateComplexMatrix5RealVector(TSdvector *x_dv)
-{
- int n, k;
- double *v;
- //
- MKL_Complex16 *Z = NULL;
-
- if (!x_dv) fn_DisplayError("CreateComplexMatrix5RealVector(): Input vector x_dv must be allocated memory");
- v = x_dv->v;
-
- Z = tzMalloc(n=x_dv->n, MKL_Complex16);
- for (k=n-1; k>=0; k--) {
- Z[k].real = v[k];
- Z[k].imag = 0.0;
- }
- return(Z);
-}
-
-
-static void ComplexMatrix2RealMatrix(TSdmatrix *Y_dm, MKL_Complex16 *Z)
-{
- int _k;
- double *M;
-
- if (!Y_dm) fn_DisplayError("ComplexMatrix2RealMatrix(): Output matrix Y_dm must be allocated memory");
- M = Y_dm->M;
-
- for (_k=Y_dm->nrows*Y_dm->ncols-1; _k>=0; _k--) M[_k] = Z[_k].real;
- Y_dm->flag = M_GE;
-}
-
-
-static void ComplexMatrix2RealVector(TSdvector *y_dv, MKL_Complex16 *Z)
-{
- int _k;
- double *v;
-
- if (!y_dv) fn_DisplayError("ComplexMatrix2RealVector(): Output matrix y_dv must be allocated memory");
- v = y_dv->v;
-
- for (_k=y_dv->n-1; _k>=0; _k--) v[_k] = Z[_k].real;
- y_dv->flag = V_DEF;
-}
-
-
-static TSdzmatrix *SubComplexMatrix2Zmatrix(TSdzmatrix *X_dzm, MKL_Complex16 *Z, const int nrowsforZ, const int _m, const int _n)
-{
- //X_dzm is _m-by_n comlex types where nrowsforZ <= _m and Z is nrowsforZ-by-_n.
- int _i, _j, incz;
- double *Mreal, *Mimag;
- MKL_Complex16 *zp;
- //
- TSdzmatrix *Y_dzm = NULL;
-
- if (!X_dzm || X_dzm->real->nrows != _m || X_dzm->real->ncols != _n) {
- DestroyMatrix_dz(X_dzm); //Destroys Y_dzm if already allocated memory to accommodate a possbible change of its dimension.
- Y_dzm = CreateMatrix_dz(_m, _n);
- }
- else Y_dzm = X_dzm;
-
- Mreal = Y_dzm->real->M;
- Mimag = Y_dzm->imag->M;
- zp = Z;
- if ((incz=nrowsforZ-_m)<0) fn_DisplayError("SubComplexMatrix2ZMatrix(): Number of rows for the input complex matrix Z must be greater that of the output Z matrix Y_dzm");
-
- for (_j=0; _j<_n; _j++) {
- for (_i=0; _i<_m; _i++) {
- *Mreal++ = zp->real;
- *Mimag++ = zp->imag;
- zp++;
- }
- zp += incz;
- }
- return (Y_dzm);
-}
-
-
-static void InitializeConstantMLK_Complex16(MKL_Complex16 *x_clx, const int _n, const double c)
-{
- int _i;
-
- for (_i=_n-1; _i>=0; _i--)
- x_clx[_i].real = x_clx[_i].imag = c;
-}
-
-static void InitializeConstantDouble(double *x_p, const int _n, const double c)
-{
- int _i;
-
- for (_i=_n-1; _i>=0; _i--) x_p[_i] = c;
-}
diff --git a/CFiles/gensys01Mar06Works.h b/CFiles/gensys01Mar06Works.h
deleted file mode 100755
index e2a347772044ea29ed6f7b283448d74a7bb90c02..0000000000000000000000000000000000000000
--- a/CFiles/gensys01Mar06Works.h
+++ /dev/null
@@ -1,65 +0,0 @@
-/*******************************************************************
- * [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
- *
- * System given as
- * g0*y(t)=g1*y(t-1)+c+psi*z(t)+pi*eta(t),
- * with z an exogenous variable process and eta being endogenously determined
- * one-step-ahead expectational errors. Returned system is
- * y(t)=G1*y(t-1)+C+impact*z(t)+ywt*inv(I-fmat*inv(L))*fwt*z(t+1) .
- * If z(t) is i.i.d., the last term drops out.
- * If div or stake is omitted from argument list, a div>1 or stake>1 is calculated.
- * eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
- * existence only with not-serially correlated z(t); eu=[-2,-2] for coincident zeros.
- *
- * g0, g1: n-by-n matrices.
- * c: n-by-1 constant terms.
- * z(t): m-by-1 vector of exogenous residuals where m < n.
- * psi: n-by-m matrix.
- * eta(t): h-by-1 vector of expectational (endogenous) errors.
- * pi: n-by-h matrix.
- * div: a real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- *-------
- * G1 or Theta_dm: n-by-n matrices.
- * C: n-by-1 vector of constant terms.
- * impact: n-by-m matrix.
- * gev: n-by-2 z vector of stacked generalized eigenvalues where gev(;,2) ./ gev(:,1) = eig(g0, g1).
- * ywt: n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- * fmat: nunstab-by-nunstab z matrix where nunstab is the number of non-stable roots.
- * fwt: nunstab-by-m z matrix.
-********************************************************************/
-
-#ifndef __GENSYS_H__
- #define __GENSYS_H__
-
- #include "tzmatlab.h"
-
- #define REALSMALL 1e-7
-
- typedef struct TSgensys_tag {
- //=== Output arguments.
- TSdmatrix *Theta_dm; //n-by-n.
- TSdvector *c_dv; //n-by-1.
- TSdmatrix *Impact_dm; //n-by-m.
- TSdzmatrix *Fmat_dzm; //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- TSdzmatrix *Fwt_dzm; //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- TSdzmatrix *Ywt_dzm; //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- TSdzmatrix *Gev_dzm; //n-by-2 z matrix of possible complex numbers.
- TSivector *eu_iv; //2-by-1.
- //=== Function itself.
- void (*gensys)(struct TSgensys_tag *, void *);
- //=== Input arguments, which are all intialized to 0.0 and whose flags are set to M_GE.
- TSdmatrix *G0_dm; //n-by-n.
- TSdmatrix *G1_dm; //n-by-n.
- TSdvector *c0_dv; //n-by-1.
- TSdmatrix *Psi_dm; //n-by-m.
- TSdmatrix *Pi_dm; //n-by-k whtere k is the number of expectational errors.
- double div; //A real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- } TSgensys;
- //
- typedef void TFlinratexp(struct TSgensys_tag *, void *); //For linear rational expectations models.
-
- struct TSgensys_tag *CreateTSgensys(TFlinratexp *func, const int _n, const int _m, const int _k, const double div);
- struct TSgensys_tag *DestroyTSgensys(struct TSgensys_tag *gensys_ps);
- void gensys_sims(struct TSgensys_tag *gensys_ps, void *dummy_ps);
-#endif
-
diff --git a/CFiles/gensys08Mar06.c b/CFiles/gensys08Mar06.c
deleted file mode 100755
index f51c7c689ed749c1e1006944c17f1cf84f301445..0000000000000000000000000000000000000000
--- a/CFiles/gensys08Mar06.c
+++ /dev/null
@@ -1,1113 +0,0 @@
-/*******************************************************************
- * [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
- *
- * System given as
- * g0*y(t)=g1*y(t-1)+c+psi*z(t)+pi*eta(t),
- * with z an exogenous variable process and eta being endogenously determined
- * one-step-ahead expectational errors. Returned system is
- * y(t)=G1*y(t-1)+C+impact*z(t)+ywt*inv(I-fmat*inv(L))*fwt*z(t+1) .
- * If z(t) is i.i.d., the last term drops out.
- * If div or stake is omitted from argument list, a div>1 or stake>1 is calculated.
- * eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
- * existence only with not-serially correlated z(t); eu=[-2,-2] for coincident zeros.
- *
- * g0, g1: n-by-n matrices.
- * c: n-by-1 constant terms.
- * z(t): m-by-1 vector of exogenous residuals where m < n.
- * psi: n-by-m matrix.
- * eta(t): h-by-1 vector of expectational (endogenous) errors.
- * pi: n-by-h matrix.
- * div: a real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- *-------
- * G1 or Theta_dm: n-by-n matrices.
- * C: n-by-1 vector of constant terms.
- * impact: n-by-m matrix.
- * gev: n-by-2 z vector of stacked generalized eigenvalues where gev(;,2) ./ gev(:,1) = eig(g0, g1).
- * ywt: n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- * fmat: nunstab-by-nunstab z matrix where nunstab is the number of non-stable roots.
- * fwt: nunstab-by-m z matrix.
- *
- * 1996 MATLAB algorithm by Christopher Sims
- * 2002 Mex implementation by Iskander Karibzhanov
- * 2004 Modified to C function by Tao Zha (April), correcting a few bugs of Iskander.
- * 03/01/06 Another modification by Tao Zha, to be consistent with the CAS 3/10/04 correction.
- *
- * Note: Iskander is transforming g0 and g1 to complex matrices and uses zgges() as a qz decomposition.
- * This is really wasting efficiency. One should keep g0 and g1 as real matrices and use
- * dgges() as a qz decomposition. I don't have time to overhaul this at this point. 04/20/04, T. Zha.
- * Note: 02/22/06. I take the above note back. According to DW, it is easy to *order* the
- * the generalized eigenvalues by using the complex g0 and g1. In principle, one could
- * order the roots using the real qz decomposition on real matrices g0 and g1. But so far
- * Dan has found it a pain to do it. Perhaps we should read the MKL Lapack manual more
- * carefully at a later point.
-********************************************************************/
-#include "gensys.h"
-
-
-//----- NOTE: We can't replace MKL_Complex16 with a different name because the Intel Lapack uses MKL_Complex16.
-//----- The only way to do this is to overhaul the code and put a wrapper function on each Intel Lapack function.
-static int selctg(MKL_Complex16 *alpha, MKL_Complex16 *beta);
-static int qz(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *q, MKL_Complex16 *z, int n);
-static MKL_Complex16* CreateComplexMatrix5RealMatrix(TSdmatrix *X_dm);
-static MKL_Complex16* CreateComplexMatrix5RealVector(TSdvector *x_dv);
-static void ComplexMatrix2RealMatrix(TSdmatrix *Y_dm, MKL_Complex16 *Z);
-static void ComplexMatrix2RealVector(TSdvector *y_dv, MKL_Complex16 *Z);
-static TSdzmatrix *SubComplexMatrix2Zmatrix(TSdzmatrix *X_dzm, MKL_Complex16 *Z, const int nrowsforZ, const int _m, const int _n);
-static void copy_eigenvalues(TSdzmatrix *Gev_dzm, MKL_Complex16 *a, MKL_Complex16 *b);
-static int compute_svd(MKL_Complex16 *a, MKL_Complex16 **u, double **d, MKL_Complex16 **v, int m, int n);
-static int compute_norm(MKL_Complex16 *a, double **d, int m, int n);
-//--- 03/01/06 TZ. Commented out to be consistent with the CAS 3/10/04 correction.
-// static int compute_normx(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *zwt, MKL_Complex16 *ueta, double **normx, int nunstab, int psin, int n, int bigev);
-static void cblas_zdupe(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdscali(int n, double *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdscale(int n, double *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdpsb(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb, MKL_Complex16 *c, int ldc);
-//
-static void InitializeConstantMLK_Complex16(MKL_Complex16 *x_clx, const int _n, const double c);
-static void InitializeConstantDouble(double *x_p, const int _n, const double c);
-
-
-
-TSgensys *CreateTSgensys(TFlinratexp *func, const int _n, const int _m, const int _k, const double div)
-{
- //_n is the number of stacked variables (endogenous, Lagurangian multiplier, expected multiplier, etc.).
- //_m is the number of exogenous shocks.
- //_k is the number of expectational errors.
- //div is the dividing number to determine what constitutes an unstable root. If div<1.0, a div>1.0 is calculated mechanically.
- TSgensys *gensys_ps = tzMalloc(1, TSgensys);
-
- //=== Output arguments.
- gensys_ps->Theta_dm = CreateMatrix_lf(_n, _n); //n-by-n.
- gensys_ps->c_dv = CreateVector_lf(_n); //n-by-1.
- gensys_ps->Impact_dm = CreateMatrix_lf(_n, _m); //n-by-m.
- gensys_ps->Fmat_dzm = (TSdzmatrix *)NULL; //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- gensys_ps->Fwt_dzm = (TSdzmatrix *)NULL; //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- gensys_ps->Ywt_dzm = (TSdzmatrix *)NULL; //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- gensys_ps->Gev_dzm = CreateMatrix_dz(_n, 2); //n-by-2 z matrix of possible complex numbers.
- gensys_ps->eu_iv = CreateConstantVector_int(2, 0); //2-by-1.
-
- //=== Function itself.
- gensys_ps->gensys = func;
-
- //=== Input arguments.
- gensys_ps->G0_dm = CreateConstantMatrix_lf(_n, _n, 0.0); //n-by-n.
- gensys_ps->G0_dm->flag = M_GE;
- gensys_ps->G1_dm = CreateConstantMatrix_lf(_n, _n, 0.0); //n-by-n.
- gensys_ps->G1_dm->flag = M_GE;
- gensys_ps->c0_dv = CreateConstantVector_lf(_n, 0.0); //n-by-1.
- gensys_ps->Psi_dm = CreateConstantMatrix_lf(_n, _m, 0.0); //n-by-m.
- gensys_ps->Psi_dm->flag = M_GE;
- gensys_ps->Pi_dm = CreateConstantMatrix_lf(_n, _k, 0.0); //n-by-k where k is the number of expectational errors.
- gensys_ps->Pi_dm->flag = M_GE;
- gensys_ps->div = div;
-
- return (gensys_ps);
-}
-//-------
-TSgensys *DestroyTSgensys(TSgensys *gensys_ps)
-{
- if (gensys_ps) {
- //=== Output arguments.
- DestroyMatrix_lf(gensys_ps->Theta_dm); //n-by-n.
- DestroyVector_lf(gensys_ps->c_dv); //n-by-1.
- DestroyMatrix_lf(gensys_ps->Impact_dm); //n-by-m.
- DestroyMatrix_dz(gensys_ps->Fmat_dzm); //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- DestroyMatrix_dz(gensys_ps->Fwt_dzm); //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- DestroyMatrix_dz(gensys_ps->Ywt_dzm); //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- DestroyMatrix_dz(gensys_ps->Gev_dzm); //n-by-2 z matrix of possible complex numbers.
- DestroyVector_int(gensys_ps->eu_iv); //2-by-1.
-
- //=== Input arguments.
- DestroyMatrix_lf(gensys_ps->G0_dm); //n-by-n.
- DestroyMatrix_lf(gensys_ps->G1_dm); //n-by-n.
- DestroyVector_lf(gensys_ps->c0_dv); //n-by-1.
- DestroyMatrix_lf(gensys_ps->Psi_dm); //n-by-m.
- DestroyMatrix_lf(gensys_ps->Pi_dm); //n-by-k where k is the number of expectational errors.
-
- free(gensys_ps);
-
- return ((TSgensys *)NULL);
- }
- else return (gensys_ps);
-}
-
-
-//--------------------------- For the function gensys_sims() ------------------------------------
-static int fixdiv = 1, zxz = 0;
-static double stake = 1.01;
-static int nunstab = 0;
-static MKL_Complex16 one, minusone, zero;
-
-/* [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensysmkl(g0,g1,c,psi,pi,stake) */
-//void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
-void gensys_sims(TSgensys *gensys_ps, void *dummy_ps)
-{
- int tmpi;
- int n, psin, pin, nsquare, md, md1, i, bigev, bigev1; //mds, bigevs, //03/01/06 TZ. Commented out to be consistent with the CAS 3/10/04 correction.
- int *eu;
- int exist = 0, existx = 0, unique = 0;
- //=== Memory will be allocated to the following.
- double *deta = NULL, *deta1 = NULL, *norm = NULL; //*normx = NULL, *dz = NULL, //03/01/06 TZ. Commented out to be consistent with the CAS 3/10/04 correction.
- MKL_Complex16 *a = NULL, *b = NULL, *q = NULL, *z = NULL, *pi = NULL, *psi = NULL;
- MKL_Complex16 *tmat = NULL, *g0 = NULL, *g1 = NULL, *dummy = NULL, *tmatq = NULL, *c = NULL, *impact = NULL, *ab = NULL;
- MKL_Complex16 *fmat = NULL, *fwt = NULL, *ywt = NULL;
- MKL_Complex16 *etawt = NULL, *ueta = NULL, *veta = NULL, *etawt1 = NULL, *ueta1 = NULL, *veta1 = NULL;
- // *uz = NULL, *vz = NULL, *zwt = NULL, //03/01/06 TZ. Commented out to be consistent with the CAS 3/10/04 correction.
- //--- Dimensions.
- n = gensys_ps->G0_dm->nrows;
- psin = gensys_ps->Psi_dm->ncols;
- pin = gensys_ps->Pi_dm->ncols;
- //--- Pointer.
- eu = gensys_ps->eu_iv->v;
-
-
- //=== [a b q z]=qz(g0,g1);
- a = CreateComplexMatrix5RealMatrix(gensys_ps->G0_dm);
- b = CreateComplexMatrix5RealMatrix(gensys_ps->G1_dm);
- q = tzMalloc(nsquare=square(n), MKL_Complex16);
- z = tzMalloc(nsquare, MKL_Complex16);
- InitializeConstantMLK_Complex16(q, nsquare, 0.0);
- InitializeConstantMLK_Complex16(z, nsquare, 0.0);
-
- fixdiv = (gensys_ps->div < 1.0);
- stake = fixdiv ? 1.01 : gensys_ps->div;
- nunstab = 0;
- zxz = 0;
-
- if (qz(a, b, q, z, n)) {
- printf("WARNING: QZ factorization failed.\n");
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
-
- nunstab /= 2;
-
- if (zxz) {
- printf("WARNING: Coincident zeros. Indeterminacy and/or nonexistence.\n");
- eu[0] = eu[1] = -2;
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- copy_eigenvalues(gensys_ps->Gev_dzm, a, b);
-
- one.real = 1.0;
- one.imag = 0.0;
-
- minusone.real = -1.0;
- minusone.imag = 0.0;
-
- zero.real = 0.0;
- zero.imag = 0.0;
-
- pi = CreateComplexMatrix5RealMatrix(gensys_ps->Pi_dm);
- etawt = tzMalloc(tmpi=nunstab*pin, MKL_Complex16);
- InitializeConstantMLK_Complex16(etawt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, pin, n,
- &one, q+n*(n-nunstab), n, pi, n, &zero, etawt, nunstab);
- if (compute_svd(etawt, &ueta, &deta, &veta, nunstab, pin)) {
- //Memory is now allocated to ueta, deta, and veta.
-
- //=== DDDDDebugging
- fprintf(FPTR_DEBUG, "Aind=[\n");
- WriteMatrix(FPTR_DEBUG, gensys_ps->G0_dm, " %.16e ");
- fprintf(FPTR_DEBUG, "];\n");
- fprintf(FPTR_DEBUG, "Bind=[\n");
- WriteMatrix(FPTR_DEBUG, gensys_ps->G1_dm, " %.16e ");
- fprintf(FPTR_DEBUG, "];\n");
- fprintf(FPTR_DEBUG, "Consterm=[\n");
- WriteVector(FPTR_DEBUG, gensys_ps->c0_dv, " %.16e ");
- fprintf(FPTR_DEBUG, "]';\n");
- fprintf(FPTR_DEBUG, "gUpsiloneind=[\n");
- WriteMatrix(FPTR_DEBUG, gensys_ps->Psi_dm, " %.16e ");
- fprintf(FPTR_DEBUG, "];\n");
- fprintf(FPTR_DEBUG, "gUpsilonxind=[\n");
- WriteMatrix(FPTR_DEBUG, gensys_ps->Pi_dm, " %.16e ");
- fprintf(FPTR_DEBUG, "];\n");
- fflush(FPTR_DEBUG);
-
-
- printf("WARNING: SVD failed.\n");
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(etawt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(etawt);
- md = nunstab<pin?nunstab:pin;
- bigev = md;
- for (i=0; i<md; i++)
- if (deta[i]<=REALSMALL) {
- bigev=i;
- break;
- }
- //------ 03/01/06 TZ: corrected code by CAS, 3/10/04.
- eu[0] = (bigev >= nunstab);
- //---------------------------------
- // ueta = nunstab x bigev
- // deta = bigev x 1
- // veta = bigev x pin, ldveta = md
- // uz = nunstab x bigevs
- // dz = bigevs x 1
- // vz = bigevs x psin, ldvz = mds
- //---------------------------------
-
- //====== 03/01/06 TZ: the following note is added by CAS 3/10/04.
- //------ Code below allowed "existence" in cases where the initial lagged state was free to take on values
- //------ inconsistent with existence, so long as the state could w.p.1 remain consistent with a stable solution
- //------ if its initial lagged value was consistent with a stable solution. This is a mistake, though perhaps there
- //------ are situations where we would like to know that this "existence for restricted initial state" situation holds.
- // psi = CreateComplexMatrix5RealMatrix(gensys_ps->Psi_dm);
- // zwt = tzMalloc(tmpi=nunstab*psin, MKL_Complex16);
- // InitializeConstantMLK_Complex16(zwt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- // cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, psin, n,
- // &one, q+n*(n-nunstab), n, psi, n, &zero, zwt, nunstab);
- // if (compute_svd(zwt, &uz, &dz, &vz, nunstab, psin)) {
- // //Memory is now allocated to uz, dz, and vz.
- // printf("WARNING: SV decomposition failed.\n");
- // tzDestroy(ueta);
- // tzDestroy(deta);
- // tzDestroy(veta);
- // tzDestroy(uz);
- // tzDestroy(dz);
- // tzDestroy(vz);
- // tzDestroy(zwt);
- // tzDestroy(a);
- // tzDestroy(b);
- // tzDestroy(q);
- // tzDestroy(z);
- // return;
- // }
- // tzDestroy(vz);
- // mds = nunstab<psin?nunstab:psin;
- // bigevs = mds;
- // for (i=0; i<mds; i++)
- // if (dz[i]<=REALSMALL) {
- // bigevs=i;
- // break;
- // }
- // tzDestroy(dz);
- //
- // if (!bigevs) {
- // exist = 1;
- // existx = 1;
- // } else {
- // /* uz-ueta*ueta'*uz */
- // MKL_Complex16 *tmp = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- // InitializeConstantMLK_Complex16(tmp, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- // cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, nunstab, nunstab,
- // bigev, &one, ueta, nunstab, ueta, nunstab, &zero, tmp, nunstab);
- // cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, nunstab,
- // bigevs, &minusone, tmp, nunstab, uz, nunstab, &one, uz, nunstab);
- // tzDestroy(tmp);
- // if (compute_norm(uz, &norm, nunstab, bigevs)) {
- // //Memory is now allocated to norm.
- // printf("WARNING: SVD failed.\n");
- // tzDestroy(norm);
- // tzDestroy(ueta);
- // tzDestroy(deta);
- // tzDestroy(veta);
- // tzDestroy(uz);
- // tzDestroy(zwt);
- // tzDestroy(a);
- // tzDestroy(b);
- // tzDestroy(q);
- // tzDestroy(z);
- // return;
- // }
- // exist = *norm < REALSMALL*n;
- // tzDestroy(norm);
- // if (compute_normx(a, b, zwt, ueta, &normx, nunstab, psin, n, bigev)) {
- // //If 0, memory is now allocated to normx; otherwise, normx is destroyed within the function compute_normx().
- // tzDestroy(ueta);
- // tzDestroy(deta);
- // tzDestroy(veta);
- // tzDestroy(uz);
- // tzDestroy(zwt);
- // tzDestroy(a);
- // tzDestroy(b);
- // tzDestroy(q);
- // tzDestroy(z);
- // return;
- // }
- // existx = *normx < REALSMALL*n;
- // tzDestroy(normx);
- // }
- //
- // tzDestroy(uz);
- // tzDestroy(zwt);
-
- //---------------------------------------------------------------------------
- // Note that existence and uniqueness are not just matters of comparing
- // numbers of roots and numbers of endogenous errors. These counts are
- // reported below because usually they point to the source of the problem.
- //---------------------------------------------------------------------------
- etawt1 = tzMalloc(tmpi=(n-nunstab)*pin, MKL_Complex16);
- InitializeConstantMLK_Complex16(etawt1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, n-nunstab, pin, n,
- &one, q, n, pi, n, &zero, etawt1, n-nunstab);
- tzDestroy(pi);
- if (compute_svd(etawt1, &ueta1, &deta1, &veta1, n-nunstab, pin)) {
- //Memory is now allocated to ueta1, deta1, and veta1.
- printf("WARNING: SVD failed for compute_svd().\n");
- tzDestroy(ueta1);
- tzDestroy(deta1);
- tzDestroy(veta1);
- tzDestroy(etawt1);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(etawt1);
- md1 = n-nunstab<pin?n-nunstab:pin;
- bigev1 = md1;
- for (i=0; i<md1; i++)
- if (deta1[i]<=REALSMALL) {
- bigev1=i;
- break;
- }
-
- //====== 03/01/06 TZ: the following is commented out by CAS 3/10/04.
- // if (existx || !nunstab) {
- // //=== Solution exists.
- // eu[0] = 1;
- // } else {
- // if (exist) {
- // printf("WARNING: Solution exists for unforecastable z only\n");
- // eu[0] = -1;
- // } /* else
- // mexPrintf("No solution. %d unstable roots. %d endog errors.\n",nunstab,bigev1); */
- // /* mexPrintf("Generalized eigenvalues\n");
- // mexCallMATLAB(0,NULL,1, &plhs[6], "disp"); */
- // }
-
-
- //-------------------------------
- // ueta1 = n-nunstab x bigev1
- // deta1 = bigev1 x 1
- // veta1 = bigev1 x pin, ldveta1 = md1
- //-------------------------------
- if (!bigev1)
- unique = 1;
- else {
- // veta1-veta1*veta*veta'
- // veta = bigev x pin, ldveta1 = md
- // veta1 = bigev1 x pin, ldveta1 = md1
- MKL_Complex16 *tmp = tzMalloc(pin*pin, MKL_Complex16);
- MKL_Complex16 *veta1_copy = tzMalloc(pin*bigev1, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, pin*pin, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- InitializeConstantMLK_Complex16(veta1_copy, pin*bigev1, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, pin, pin,
- bigev, &one, veta, md, veta, md, &zero, tmp, pin);
- cblas_zdupe(bigev1,pin,veta1,md1,veta1_copy,bigev1);
- cblas_zhemm(CblasColMajor, CblasRight, CblasUpper, bigev1, pin,
- &minusone, tmp, pin, veta1_copy, bigev1, &one, veta1_copy, bigev1);
- tzDestroy(tmp);
- if (compute_norm(veta1_copy, &norm, bigev1, pin)) {
- //Memory is now allocated to norm.
- printf("WARNING: SVD failed.\n");
- tzDestroy(norm);
- tzDestroy(ueta1);
- tzDestroy(deta1);
- tzDestroy(veta1);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(veta1_copy);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(veta1_copy);
- unique = *norm < REALSMALL*n;
- tzDestroy(norm);
- }
- if (unique) {
- //=== Unique solution.
- eu[1] = 1;
- } else {
- eu[1] = 0;
- printf("WARNING: Indeterminacy. %d loose endog errors with eu being [%d, %d].\n",bigev1-bigev, eu[0], eu[1]);
- //printf("WARNING: Indeterminacy. %d loose endog errors with eu being [%g, %g].\n",bigev1-bigev, gensys_ps->eu_dv->v[0], gensys_ps->eu_dv->v[1]);
- //printf("WARNING: Indeterminacy. %d loose endog errors.\n",bigev1-bigev);
- }
-
- cblas_zdscali(pin,deta,bigev,veta,md); /* veta' = deta\veta' */
- tzDestroy(deta);
- cblas_zdscale(pin,deta1,bigev1,veta1,md1); /* veta1' = deta1*veta1' */
- tzDestroy(deta1);
- etawt = tzMalloc(tmpi=nunstab*pin, MKL_Complex16); /* etawt = ueta*veta' */
- InitializeConstantMLK_Complex16(etawt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, nunstab, pin, bigev,
- &one, ueta, nunstab, veta, md, &zero, etawt, nunstab);
- tzDestroy(ueta);
- tzDestroy(veta);
- etawt1 = tzMalloc(tmpi=(n-nunstab)*pin, MKL_Complex16); /* etawt1 = ueta1*veta1' */
- InitializeConstantMLK_Complex16(etawt1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, pin, bigev1,
- &one, ueta1, n-nunstab, veta1, md1, &zero, etawt1, n-nunstab);
- tzDestroy(ueta1);
- tzDestroy(veta1);
- tmat = tzMalloc(tmpi=(n-nunstab)*nunstab, MKL_Complex16); /* tmat = etawt1*etawt' */
- InitializeConstantMLK_Complex16(tmat, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n-nunstab, nunstab, pin,
- &one, etawt1, n-nunstab, etawt, nunstab, &zero, tmat, n-nunstab);
- tzDestroy(etawt1);
- tzDestroy(etawt);
-
- g0 = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(g0, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdupe(n-nunstab, n, a, n, g0, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, nunstab, nunstab,
- &minusone, tmat, n-nunstab, a+(n-nunstab)*(n+1), n, &one, g0+(n-nunstab)*n, n);
- cblas_zcopy(nunstab, &one, 0, g0+(n-nunstab)*(n+1), n+1);
-
- g1 = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(g1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdupe(n-nunstab, n, b, n, g1, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, nunstab, nunstab,
- &minusone, tmat, n-nunstab, b+(n-nunstab)*(n+1), n, &one, g1+(n-nunstab)*n, n);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, n, &one, g0, n, g1, n);
- dummy = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, n, n, &one, z, n, g1, n, &zero, dummy, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n, n, n, &one, dummy, n, z, n, &zero, g1, n);
- tzDestroy(dummy);
- ComplexMatrix2RealMatrix(gensys_ps->Theta_dm, g1); //Output.
- tzDestroy(g1);
-
- tmatq = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmatq, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zcopy(n*n, q, 1, tmatq, 1);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n, n-nunstab, nunstab,
- &minusone, tmatq+(n-nunstab)*n, n, tmat, n-nunstab, &one, tmatq, n);
- tzDestroy(tmat);
-
- ab = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(ab, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdpsb(nunstab, nunstab, a+(n-nunstab)*(n+1), n, b+(n-nunstab)*(n+1), n, ab, nunstab);
- cblas_ztrsm(CblasColMajor, CblasRight, CblasUpper, CblasConjTrans, CblasNonUnit,
- n, nunstab, &one, ab, nunstab, tmatq+(n-nunstab)*n, n);
- tzDestroy(ab);
-
- //c = mat2mkl(prhs[2],n,1);
- c = CreateComplexMatrix5RealVector(gensys_ps->c0_dv);
- dummy = tzMalloc(gensys_ps->c0_dv->n, MKL_Complex16);
- //$$$$$$$ The following is Iskander's fatal code error. One cannot use c in the two different places; otherwise, it makes c be zero completely!
- // cblas_zgemv(CblasColMajor, CblasConjTrans, n, n, &one, tmatq, n, c, 1, &zero, c, 1);
- // cblas_ztrsv(CblasColMajor, CblasUpper, CblasNoTrans, CblasNonUnit, n, g0, n, c, 1);
- // cblas_zgemv(CblasColMajor, CblasNoTrans, n, n, &one, z, n, c, 1, &zero, c, 1);
- cblas_zgemv(CblasColMajor, CblasConjTrans, n, n, &one, tmatq, n, c, 1, &zero, dummy, 1);
- cblas_ztrsv(CblasColMajor, CblasUpper, CblasNoTrans, CblasNonUnit, n, g0, n, dummy, 1);
- cblas_zgemv(CblasColMajor, CblasNoTrans, n, n, &one, z, n, dummy, 1, &zero, c, 1);
- //plhs[1] = mkl2mat_real(c, n, n, 1);
- ComplexMatrix2RealVector(gensys_ps->c_dv, c); //Output.
- tzDestroy(c);
- tzDestroy(dummy);
-
- impact = tzMalloc(tmpi=n*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(impact, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- psi = CreateComplexMatrix5RealMatrix(gensys_ps->Psi_dm); //03/01/06 TZ. Added to be consistent with the CAS 3/10/04 correction.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, n-nunstab, psin, n,
- &one, tmatq, n, psi, n, &zero, impact, n);
- tzDestroy(tmatq);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, psin, &one, g0, n, impact, n);
- dummy = tzMalloc(tmpi=n*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, psin, n,
- &one, z, n, impact, n, &zero, dummy, n);
- tzDestroy(impact);
- //plhs[2] = mkl2mat_real(dummy, n, n, psin);
- ComplexMatrix2RealMatrix(gensys_ps->Impact_dm, dummy); //Output.
- tzDestroy(dummy);
-
- fmat = a+(n-nunstab)*(n+1);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, nunstab, &one, b+(n-nunstab)*(n+1), n, fmat, n);
- //plhs[3] = mkl2mat(fmat, n, nunstab, nunstab);
- gensys_ps->Fmat_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Fmat_dzm, fmat, n, nunstab, nunstab);
- tzDestroy(a);
-
- fwt = tzMalloc(tmpi=nunstab*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(fwt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_ztrsm(CblasColMajor, CblasRight, CblasUpper, CblasConjTrans, CblasNonUnit,
- n, nunstab, &one, b+(n-nunstab)*(n+1), n, q+(n-nunstab)*n, n);
- tzDestroy(b);
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, psin, n,
- &minusone, q+(n-nunstab)*n, n, psi, n, &zero, fwt, nunstab);
- tzDestroy(q);
- tzDestroy(psi);
- gensys_ps->Fwt_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Fwt_dzm, fwt, nunstab, nunstab, psin);
- tzDestroy(fwt);
-
- ywt = tzMalloc(tmpi=n*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(ywt, tmpi, 0.0);
- cblas_zcopy(nunstab, &one, 0, ywt+n-nunstab, n+1);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, nunstab, &one, g0, n, ywt, n);
- tzDestroy(g0);
- dummy = tzMalloc(tmpi=n*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, nunstab, n,
- &one, z, n, ywt, n, &zero, dummy, n);
- tzDestroy(z);
- tzDestroy(ywt);
- gensys_ps->Ywt_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Ywt_dzm, dummy, n, n, nunstab);
- tzDestroy(dummy);
-}
-
-
-/******************************************************************************
- * function selctg orders the eigenvalues so that a selected cluster of *
- * eigenvalues appears in the leading diagonal blocks of the upper *
- * quasi-triangular matrix S and the upper triangular matrix T. *
- ******************************************************************************/
-
-static int selctg(MKL_Complex16 *alpha, MKL_Complex16 *beta)
-{
- double absA = sqrt(alpha->real*alpha->real+alpha->imag*alpha->imag),
- absB = fabs(beta->real);
- if (absA) {
- double divhat = absB/absA;
- //bug detected by Vasco Curdia and Daria Finocchiaro, 2/25/2004 CAS. A root of
- //exactly 1.01 and no root between 1 and 1.02, led to div being stuck at 1.01
- //and the 1.01 root being misclassified as stable. Changing < to <= below fixes this.
- if (fixdiv && 1+REALSMALL<divhat && divhat<=stake)
- stake = (1+divhat)/2;
- }
- if (absA<REALSMALL && absB<REALSMALL)
- zxz = 1;
- if (absB>stake*absA) {
- nunstab++;
- return(0);
- } else
- return(1);
-}
-
-/******************************************************************************
- * compute for a pair of N-by-N complex nonsymmetric matrices (A,B), *
- * the generalized eigenvalues, the generalized complex Schur form (S, T), *
- * and optionally left and/or right Schur vectors (VSL and VSR) *
- ******************************************************************************/
-
-static int qz(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *q, MKL_Complex16 *z, int n)
-{
-// unsigned char msg[101];
- int sdim, lwork = -1, info = 0;
- MKL_Complex16 *alpha = tzMalloc(n,MKL_Complex16),
- *beta = tzMalloc(n,MKL_Complex16),
- *work, work1;
- double *rwork = tzMalloc(8*n, double);
- int *bwork = tzMalloc(4*n, int);
-
- /* Query zgges on the value of lwork */
- zgges("V", "V", "S", &selctg, &n, a, &n, b, &n, &sdim, alpha, beta, q,
- &n, z, &n, &work1, &lwork, rwork, bwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to the Intel MKL function zgges() has an illegal value",-info);
- tzDestroy(bwork);
- tzDestroy(rwork);
- tzDestroy(alpha);
- tzDestroy(beta);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgges("V", "V", "S", &selctg, &n, a, &n, b, &n, &sdim, alpha, beta, q,
- &n, z, &n, work, &lwork, rwork, bwork, &info);
-
- tzDestroy(work);
- tzDestroy(bwork);
- tzDestroy(rwork);
- tzDestroy(alpha);
- tzDestroy(beta);
-
- if (info < 0) {
- printf("WARNING: Input %d to the Intel MKL function zgges() has an illegal value",-info);
- return(info);
- }
-
- if (info > 0)
- if (info < n)
- printf("WARNING: The QZ iteration failed. (A,B) are not in Schur form,\n"
- "but ALPHA(j) and BETA(j) should be correct for j=%d,...,N.",info+1);
- else {
- switch (info-n) {
- case 1:
- printf("WARNING: LAPACK problem: error return from ZGGBAL");
- break;
- case 2:
- printf("WARNING: LAPACK problem: error return from ZGEQRF");
- break;
- case 3:
- printf("WARNING: LAPACK problem: error return from ZUNMQR");
- break;
- case 4:
- printf("WARNING: LAPACK problem: error return from ZUNGQR");
- break;
- case 5:
- printf("WARNING: LAPACK problem: error return from ZGGHRD");
- break;
- case 6:
- printf("WARNING: LAPACK problem: error return from ZHGEQZ (other than failed iteration)");
- break;
- case 7:
- printf("WARNING: LAPACK problem: error return from ZGGBAK (computing VSL)");
- break;
- case 8:
- printf("WARNING: LAPACK problem: error return from ZGGBAK (computing VSR)");
- break;
- case 9:
- printf("WARNING: LAPACK problem: error return from ZLASCL (various places)");
- break;
- default:
- printf("WARNING: LAPACK problem: unknown error.");
- break;
- }
- }
- return(info);
-}
-
-/*
- * Convert MATLAB complex matrix to MKL complex storage.
- *
- * Z = mat2mkl(X,ldz,ndz)
- *
- * converts MATLAB's mxArray X to MKL_Complex16 Z(ldz,ndz).
- * The parameters ldz and ndz determine the storage allocated for Z,
- * while mxGetM(X) and mxGetN(X) determine the amount of data copied.
- */
-
-//MKL_Complex16* mat2mkl(const mxArray *X, int ldz, int ndz) {
-// MKL_Complex16 *Z, *zp;
-// int m, n, incz, cmplxflag;
-// register int i, j;
-// double *xr, *xi;
-
-// Z = mxCalloc(ldz*ndz, sizeof(MKL_Complex16));
-// xr = mxGetPr(X);
-// xi = mxGetPi(X);
-// m = mxGetM(X);
-// n = mxGetN(X);
-// zp = Z;
-// incz = ldz-m;
-// cmplxflag = (xi != NULL);
-// for (j = 0; j < n; j++) {
-// if (cmplxflag) {
-// for (i = 0; i < m; i++) {
-// zp->real = *xr++;
-// zp->imag = *xi++;
-// zp++;
-// }
-// } else {
-// for (i = 0; i < m; i++) {
-// zp->real = *xr++;
-// zp++;
-// }
-// }
-// zp += incz;
-// }
-// return(Z);
-//}
-
-
-/*
- * Convert MKL complex storage to MATLAB real and imaginary parts.
- *
- * X = mkl2mat(Z,ldz,m,n)
- *
- * copies MKL_Complex16 Z to X, producing a complex mxArray
- * with mxGetM(X) = m and mxGetN(X) = n.
- */
-
-//mxArray* mkl2mat(MKL_Complex16 *Z, int ldz, int m, int n) {
-// int i, j, incz;
-// double *xr, *xi;
-// MKL_Complex16 *zp;
-// mxArray *X;
-
-// X = mxCreateDoubleMatrix(m,n,mxCOMPLEX);
-// xr = mxGetPr(X);
-// xi = mxGetPi(X);
-// zp = Z;
-// incz = ldz-m;
-// for (j = 0; j < n; j++) {
-// for (i = 0; i < m; i++) {
-// *xr++ = zp->real;
-// *xi++ = zp->imag;
-// zp++;
-// }
-// zp += incz;
-// }
-// return(X);
-//}
-
-//plhs[3] = mkl2mat(fmat, n, nunstab, nunstab)
-
-/*
- * Convert MKL complex storage to MATLAB real matrix ignoring imaginary part.
- *
- * X = mkl2mat(Z,ldz,m,n)
- *
- * copies MKL_Complex16 Z to X, producing a real mxArray
- * with mxGetM(X) = m and mxGetN(X) = n.
- */
-
-//mxArray* mkl2mat_real(MKL_Complex16 *Z, int ldz, int m, int n) {
-// int i, j, incz;
-// double *xr;
-// MKL_Complex16 *zp;
-// mxArray *X;
-
-// X = mxCreateDoubleMatrix(m,n,mxREAL);
-// xr = mxGetPr(X);
-// zp = Z;
-// incz = ldz-m;
-// for (j = 0; j < n; j++) {
-// for (i = 0; i < m; i++) {
-// *xr++ = zp->real;
-// zp++;
-// }
-// zp += incz;
-// }
-// return(X);
-//}
-
-//void copy_eigenvalues(mxArray *gev, MKL_Complex16 *a, MKL_Complex16 *b, int n) {
-// double *gevr = mxGetPr(gev),
-// *gevi = mxGetPi(gev);
-// int i;
-
-// for (i=0; i<n; i++, gevr++, gevi++, a+=n+1) {
-// *gevr = a->real;
-// *gevi = a->imag;
-// }
-
-// for (i=0; i<n; i++, gevr++, gevi++, b+=n+1) {
-// *gevr = b->real;
-// *gevi = b->imag;
-// }
-//}
-
-static void copy_eigenvalues(TSdzmatrix *Gev_dzm, MKL_Complex16 *a, MKL_Complex16 *b)
-{
- int n = Gev_dzm->real->nrows;
- double *gevr = Gev_dzm->real->M,
- *gevi = Gev_dzm->imag->M;
- int i;
-
- for (i=0; i<n; i++, gevr++, gevi++, a+=n+1) {
- *gevr = a->real;
- *gevi = a->imag;
- }
-
- for (i=0; i<n; i++, gevr++, gevi++, b+=n+1) {
- *gevr = b->real;
- *gevi = b->imag;
- }
-}
-
-
-static int compute_svd(MKL_Complex16 *x, MKL_Complex16 **u, double **d, MKL_Complex16 **v, int m, int n)
-{
- //$$$Memory allocated to u, d, and v will be destroyed outside this function.$$$
- int tmpi;
- int md = m<n?m:n, lwork = -1, info = 0;
- MKL_Complex16 *a, *work, work1;
- double *rwork = tzMalloc(5*md>1?5*md:1, double);
-
- a = tzMalloc(m*n, MKL_Complex16);
- cblas_zcopy(m*n, x, 1, a, 1);
-
- *u = tzMalloc(tmpi=m*md,MKL_Complex16);
- InitializeConstantMLK_Complex16(*u, tmpi, 0.0);
- *v = tzMalloc(tmpi=md*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(*v, tmpi, 0.0);
- *d = tzMalloc(md, double);
- InitializeConstantDouble(*d, md, 0.0);
-
- /* Query zgges on the value of lwork */
- zgesvd("S", "S", &m, &n, a, &m, *d, *u, &m, *v, &md, &work1, &lwork, rwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
- tzDestroy(rwork);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgesvd("S", "S", &m, &n, a, &m, *d, *u, &m, *v, &md, work, &lwork, rwork, &info);
-
- tzDestroy(work);
- tzDestroy(rwork);
- tzDestroy(a);
-
- if (info < 0)
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
-
- if (info > 0)
- printf("WARNING: ZBDSQR did not converge.\n%d superdiagonals of an intermediate "
- "bidiagonal form B did not converge to zero.",info);
- return(info);
-}
-
-static int compute_norm(MKL_Complex16 *a, double **d, int m, int n)
-{
- //Memory will be allocated to d, which will be destroyed outside this function.
- int md = m<n?m:n, lwork = -1, info = 0;
- MKL_Complex16 *work = NULL, work1;
- double *rwork = tzMalloc(5*md>1?5*md:1, double);
-
- *d = tzMalloc(md, double);
-
- /* Query zgges on the value of lwork */
- zgesvd("N", "N", &m, &n, a, &m, *d, NULL, &m, NULL, &md, &work1, &lwork, rwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
- tzDestroy(rwork);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgesvd("N", "N", &m, &n, a, &m, *d, NULL, &m, NULL, &md, work, &lwork, rwork, &info);
-
- tzDestroy(work);
- tzDestroy(rwork);
-
- if (info < 0)
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
-
- if (info > 0)
- printf("WARNING: ZBDSQR() in Intel MKL did not converge.\n%d superdiagonals of an intermediate "
- "bidiagonal form B did not converge to zero.",info);
-
- return(info);
-}
-
-
-//======= 03/01/06 TZ. Commented out to be consistent with the CAS 3/10/04 correction. =======//
-//static int compute_normx(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *zwt, MKL_Complex16 *ueta, double **normx, int nunstab, int psin, int n, int bigev)
-//{
-// //Memory is allocated to normx, which will be freed outside this function.
-// int tmpi;
-// int info = 0, i, bigevs;
-// //
-// MKL_Complex16 *M = NULL, *zwtx = NULL, *ux = NULL, *vx = NULL, *tmp = NULL;
-// double *dx = NULL;
-
-
-// a += (n+1)*(n-nunstab);
-// b += (n+1)*(n-nunstab);
-// cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
-// nunstab, psin, &one, b, n, zwt, nunstab);
-// M = tzMalloc(nunstab*nunstab, MKL_Complex16);
-// cblas_zdupe(nunstab, nunstab, a, n, M, nunstab);
-// cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
-// nunstab, nunstab, &one, b, n, M, nunstab);
-
-// zwtx = tzMalloc(nunstab*nunstab*psin, MKL_Complex16);
-// cblas_zcopy(nunstab*psin, zwt, 1, zwtx, 1);
-// for (i=1; i<nunstab; i++) {
-// cblas_ztrmm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, nunstab, psin*i, &one, M, nunstab, zwtx, nunstab);
-// cblas_zcopy(nunstab*psin, zwt, 1, zwtx+nunstab*psin*i, 1);
-// }
-// tzDestroy(M);
-// cblas_ztrmm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, nunstab, nunstab*psin, &one, b, n, zwtx, nunstab);
-// info = compute_svd(zwtx, &ux, &dx, &vx, nunstab, nunstab*psin); //Memory is allocated to ux, dx, and vx.
-// tzDestroy(vx);
-// tzDestroy(zwtx);
-// if (info) {
-// printf("WARNING: SVD failed.\n");
-// tzDestroy(ux);
-// tzDestroy(dx);
-// return(info);
-// }
-// bigevs = nunstab;
-// for (i=0; i<nunstab; i++)
-// if (dx[i]<=REALSMALL) {
-// bigevs = i;
-// break;
-// }
-// tzDestroy(dx);
-// /* ux-ueta*ueta'*ux */
-// tmp = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
-// InitializeConstantMLK_Complex16(tmp, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
-// cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, nunstab, nunstab,
-// bigev, &one, ueta, nunstab, ueta, nunstab, &zero, tmp, nunstab);
-// cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, nunstab,
-// bigevs, &minusone, tmp, nunstab, ux, nunstab, &one, ux, nunstab);
-// tzDestroy(tmp);
-// info = compute_norm(ux, normx, nunstab, bigevs); //Memory is allocated to normx.
-// if (info) {
-// printf("WARNING: SVD failed.\n");
-// tzDestroy(normx);
-// tzDestroy(ux);
-// return(info);
-// }
-// tzDestroy(ux);
-// return(info);
-//}
-
-static void cblas_zdupe(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<m; i++, a++, b++)
- cblas_zcopy(n, a, lda, b, ldb);
-}
-
-static void cblas_zdscali(int n, double *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<lda; i++, a++, b++)
- cblas_zdscal(n, 1.0/(*a), b, ldb);
-}
-
-static void cblas_zdscale(int n, double *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<lda; i++, a++, b++)
- cblas_zdscal(n, *a, b, ldb);
-}
-
-static void cblas_zdpsb(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb, MKL_Complex16 *c, int ldc)
-{
- int i;
- cblas_zdupe(m, n, a, lda, c, ldc);
- for (i=0; i<m; i++, b++, c++)
- cblas_zaxpy(n, &minusone, b, ldb, c, ldc);
-}
-
-
-static MKL_Complex16* CreateComplexMatrix5RealMatrix(TSdmatrix *X_dm)
-{
- int mn, k;
- double *M;
- //
- MKL_Complex16 *Z = NULL;
-
- if (!X_dm) fn_DisplayError("CreateComplexMatrix5RealMatrix(): Input matrix X_dm must be allocated memory");
- M = X_dm->M;
-
- Z = tzMalloc(mn=X_dm->nrows*X_dm->ncols, MKL_Complex16);
- for (k=mn-1; k>=0; k--) {
- Z[k].real = M[k];
- Z[k].imag = 0.0;
- }
- return(Z);
-}
-
-static MKL_Complex16* CreateComplexMatrix5RealVector(TSdvector *x_dv)
-{
- int n, k;
- double *v;
- //
- MKL_Complex16 *Z = NULL;
-
- if (!x_dv) fn_DisplayError("CreateComplexMatrix5RealVector(): Input vector x_dv must be allocated memory");
- v = x_dv->v;
-
- Z = tzMalloc(n=x_dv->n, MKL_Complex16);
- for (k=n-1; k>=0; k--) {
- Z[k].real = v[k];
- Z[k].imag = 0.0;
- }
- return(Z);
-}
-
-
-static void ComplexMatrix2RealMatrix(TSdmatrix *Y_dm, MKL_Complex16 *Z)
-{
- int _k;
- double *M;
-
- if (!Y_dm) fn_DisplayError("ComplexMatrix2RealMatrix(): Output matrix Y_dm must be allocated memory");
- M = Y_dm->M;
-
- for (_k=Y_dm->nrows*Y_dm->ncols-1; _k>=0; _k--) M[_k] = Z[_k].real;
- Y_dm->flag = M_GE;
-}
-
-
-static void ComplexMatrix2RealVector(TSdvector *y_dv, MKL_Complex16 *Z)
-{
- int _k;
- double *v;
-
- if (!y_dv) fn_DisplayError("ComplexMatrix2RealVector(): Output matrix y_dv must be allocated memory");
- v = y_dv->v;
-
- for (_k=y_dv->n-1; _k>=0; _k--) v[_k] = Z[_k].real;
- y_dv->flag = V_DEF;
-}
-
-
-static TSdzmatrix *SubComplexMatrix2Zmatrix(TSdzmatrix *X_dzm, MKL_Complex16 *Z, const int nrowsforZ, const int _m, const int _n)
-{
- //X_dzm is _m-by_n comlex types where nrowsforZ <= _m and Z is nrowsforZ-by-_n.
- int _i, _j, incz;
- double *Mreal, *Mimag;
- MKL_Complex16 *zp;
- //
- TSdzmatrix *Y_dzm = NULL;
-
- if (!X_dzm || X_dzm->real->nrows != _m || X_dzm->real->ncols != _n) {
- DestroyMatrix_dz(X_dzm); //Destroys Y_dzm if already allocated memory to accommodate a possbible change of its dimension.
- Y_dzm = CreateMatrix_dz(_m, _n);
- }
- else Y_dzm = X_dzm;
-
- Mreal = Y_dzm->real->M;
- Mimag = Y_dzm->imag->M;
- zp = Z;
- if ((incz=nrowsforZ-_m)<0) fn_DisplayError("SubComplexMatrix2ZMatrix(): Number of rows for the input complex matrix Z must be greater that of the output Z matrix Y_dzm");
-
- for (_j=0; _j<_n; _j++) {
- for (_i=0; _i<_m; _i++) {
- *Mreal++ = zp->real;
- *Mimag++ = zp->imag;
- zp++;
- }
- zp += incz;
- }
- return (Y_dzm);
-}
-
-
-static void InitializeConstantMLK_Complex16(MKL_Complex16 *x_clx, const int _n, const double c)
-{
- int _i;
-
- for (_i=_n-1; _i>=0; _i--)
- x_clx[_i].real = x_clx[_i].imag = c;
-}
-
-static void InitializeConstantDouble(double *x_p, const int _n, const double c)
-{
- int _i;
-
- for (_i=_n-1; _i>=0; _i--) x_p[_i] = c;
-}
diff --git a/CFiles/gensysOldVersionWorks.c b/CFiles/gensysOldVersionWorks.c
deleted file mode 100755
index c78aa433f315a559198ce5943669d28beb5939b8..0000000000000000000000000000000000000000
--- a/CFiles/gensysOldVersionWorks.c
+++ /dev/null
@@ -1,1114 +0,0 @@
-/*******************************************************************
- * [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
- *
- * System given as
- * g0*y(t)=g1*y(t-1)+c+psi*z(t)+pi*eta(t),
- * with z an exogenous variable process and eta being endogenously determined
- * one-step-ahead expectational errors. Returned system is
- * y(t)=G1*y(t-1)+C+impact*z(t)+ywt*inv(I-fmat*inv(L))*fwt*z(t+1) .
- * If z(t) is i.i.d., the last term drops out.
- * If div or stake is omitted from argument list, a div>1 or stake>1 is calculated.
- * eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
- * existence only with not-serially correlated z(t); eu=[-2,-2] for coincident zeros.
- *
- * g0, g1: n-by-n matrices.
- * c: n-by-1 constant terms.
- * z(t): m-by-1 vector of exogenous residuals where m < n.
- * psi: n-by-m matrix.
- * eta(t): h-by-1 vector of expectational (endogenous) errors.
- * pi: n-by-h matrix.
- * div: a real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- *-------
- * G1 or Theta_dm: n-by-n matrices.
- * C: n-by-1 vector of constant terms.
- * impact: n-by-m matrix.
- * gev: n-by-2 z vector of stacked generalized eigenvalues where gev(;,2) ./ gev(:,1) = eig(g0, g1).
- * ywt: n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- * fmat: nunstab-by-nunstab z matrix where nunstab is the number of non-stable roots.
- * fwt: nunstab-by-m z matrix.
- *
- * 1996 MATLAB algorithm by Christopher Sims
- * 2002 Mex implementation by Iskander Karibzhanov
- * 2004 Modified to C function by Tao Zha (April)
- *
- * Note: Iskander is transforming g0 and g1 to complex matrices and uses zgges() as a qz decomposition.
- * This is really wasting efficiency. One should keep g0 and g1 as real matrices and use
- * dgges() as a qz decomposition. I don't have time to overhaul this at this point. 04/20/04, T. Zha.
- * Note: 02/22/06. I take the above note back. According to DW, it is easy to order the
- * the generalized eigenvalues by using the complex g0 and g1. In principle, one could
- * order the roots using the real qz decomposition on real matrices g0 and g1. But so far
- * Dan has found it a pain to do it. Perhaps we should read the MKL Lapack manual more
- * carefully at a later point.
-********************************************************************/
-
-#include "gensys.h"
-
-
-//----- NOTE: We can't replace MKL_Complex16 with a different name because the Intel Lapack uses MKL_Complex16.
-//----- The only way to do this is to overhaul the code and put a wrapper function on each Intel Lapack function.
-static int selctg(MKL_Complex16 *alpha, MKL_Complex16 *beta);
-static int qz(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *q, MKL_Complex16 *z, int n);
-static MKL_Complex16* CreateComplexMatrix5RealMatrix(TSdmatrix *X_dm);
-static MKL_Complex16* CreateComplexMatrix5RealVector(TSdvector *x_dv);
-static void ComplexMatrix2RealMatrix(TSdmatrix *Y_dm, MKL_Complex16 *Z);
-static void ComplexMatrix2RealVector(TSdvector *y_dv, MKL_Complex16 *Z);
-static TSdzmatrix *SubComplexMatrix2Zmatrix(TSdzmatrix *X_dzm, MKL_Complex16 *Z, const int nrowsforZ, const int _m, const int _n);
-static void copy_eigenvalues(TSdzmatrix *Gev_dzm, MKL_Complex16 *a, MKL_Complex16 *b);
-static int compute_svd(MKL_Complex16 *a, MKL_Complex16 **u, double **d, MKL_Complex16 **v, int m, int n);
-static int compute_norm(MKL_Complex16 *a, double **d, int m, int n);
-//
-static int compute_normx(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *zwt, MKL_Complex16 *ueta, double **normx, int nunstab, int psin, int n, int bigev);
-static void cblas_zdupe(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdscali(int n, double *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdscale(int n, double *a, int lda, MKL_Complex16 *b, int ldb);
-static void cblas_zdpsb(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb, MKL_Complex16 *c, int ldc);
-//
-static void InitializeConstantMLK_Complex16(MKL_Complex16 *x_clx, const int _n, const double c);
-static void InitializeConstantDouble(double *x_p, const int _n, const double c);
-
-
-
-TSgensys *CreateTSgensys(TFlinratexp *func, const int _n, const int _m, const int _k, const double div)
-{
- //_n is the number of stacked variables (endogenous, Lagurangian multiplier, expected multiplier, etc.).
- //_m is the number of exogenous shocks.
- //_k is the number of expectational errors.
- //div is the dividing number to determine what constitutes an unstable root. If div<1.0, a div>1.0 is calculated mechanically.
- TSgensys *gensys_ps = tzMalloc(1, TSgensys);
-
- //=== Output arguments.
- gensys_ps->Theta_dm = CreateMatrix_lf(_n, _n); //n-by-n.
- gensys_ps->c_dv = CreateVector_lf(_n); //n-by-1.
- gensys_ps->Impact_dm = CreateMatrix_lf(_n, _m); //n-by-m.
- gensys_ps->Fmat_dzm = (TSdzmatrix *)NULL; //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- gensys_ps->Fwt_dzm = (TSdzmatrix *)NULL; //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- gensys_ps->Ywt_dzm = (TSdzmatrix *)NULL; //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- gensys_ps->Gev_dzm = CreateMatrix_dz(_n, 2); //n-by-2 z matrix of possible complex numbers.
- gensys_ps->eu_dv = CreateConstantVector_lf(2, 0.0); //2-by-1.
-
- //=== Function itself.
- gensys_ps->gensys = func;
-
- //=== Input arguments.
- gensys_ps->G0_dm = CreateConstantMatrix_lf(_n, _n, 0.0); //n-by-n.
- gensys_ps->G0_dm->flag = M_GE;
- gensys_ps->G1_dm = CreateConstantMatrix_lf(_n, _n, 0.0); //n-by-n.
- gensys_ps->G1_dm->flag = M_GE;
- gensys_ps->c0_dv = CreateConstantVector_lf(_n, 0.0); //n-by-1.
- gensys_ps->Psi_dm = CreateConstantMatrix_lf(_n, _m, 0.0); //n-by-m.
- gensys_ps->Psi_dm->flag = M_GE;
- gensys_ps->Pi_dm = CreateConstantMatrix_lf(_n, _k, 0.0); //n-by-k where k is the number of expectational errors.
- gensys_ps->Pi_dm->flag = M_GE;
- gensys_ps->div = div;
-
- return (gensys_ps);
-}
-//-------
-TSgensys *DestroyTSgensys(TSgensys *gensys_ps)
-{
- if (gensys_ps) {
- //=== Output arguments.
- DestroyMatrix_lf(gensys_ps->Theta_dm); //n-by-n.
- DestroyVector_lf(gensys_ps->c_dv); //n-by-1.
- DestroyMatrix_lf(gensys_ps->Impact_dm); //n-by-m.
- DestroyMatrix_dz(gensys_ps->Fmat_dzm); //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- DestroyMatrix_dz(gensys_ps->Fwt_dzm); //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- DestroyMatrix_dz(gensys_ps->Ywt_dzm); //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- DestroyMatrix_dz(gensys_ps->Gev_dzm); //n-by-2 z matrix of possible complex numbers.
- DestroyVector_lf(gensys_ps->eu_dv); //2-by-1.
-
- //=== Input arguments.
- DestroyMatrix_lf(gensys_ps->G0_dm); //n-by-n.
- DestroyMatrix_lf(gensys_ps->G1_dm); //n-by-n.
- DestroyVector_lf(gensys_ps->c0_dv); //n-by-1.
- DestroyMatrix_lf(gensys_ps->Psi_dm); //n-by-m.
- DestroyMatrix_lf(gensys_ps->Pi_dm); //n-by-k where k is the number of expectational errors.
-
- free(gensys_ps);
-
- return ((TSgensys *)NULL);
- }
- else return (gensys_ps);
-}
-
-
-//--------------------------- For the function gensys_sims() ------------------------------------
-static int fixdiv = 1, zxz = 0;
-static double stake = 1.01;
-static int nunstab = 0;
-static MKL_Complex16 one, minusone, zero;
-
-/* [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensysmkl(g0,g1,c,psi,pi,stake) */
-//void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
-void gensys_sims(TSgensys *gensys_ps, void *dummy_ps)
-{
- int tmpi;
- int n, psin, pin, nsquare, md, mds, md1, i, bigev, bigevs, bigev1;
- double *eu;
- int exist = 0, existx = 0, unique = 0;
- //=== Memory will be allocated to the following.
- double *deta = NULL, *dz = NULL, *normx = NULL, *deta1 = NULL, *norm = NULL;
- MKL_Complex16 *a = NULL, *b = NULL, *q = NULL, *z = NULL, *pi = NULL, *psi = NULL;
- MKL_Complex16 *tmat = NULL, *g0 = NULL, *g1 = NULL, *dummy = NULL, *tmatq = NULL, *c = NULL, *impact = NULL, *ab = NULL;
- MKL_Complex16 *fmat = NULL, *fwt = NULL, *ywt = NULL;
- MKL_Complex16 *etawt = NULL, *ueta = NULL, *veta = NULL, *zwt = NULL, *uz = NULL, *vz = NULL, *etawt1 = NULL, *ueta1 = NULL, *veta1 = NULL;
-
-// /* Check for proper number of input and output arguments */
-// if (nrhs < 5 || nrhs > 6)
-// mexErrMsgTxt("Five or six input arguments required.\n");
-// if (nlhs != 8)
-// mexErrMsgTxt("Eight output arguments required.\n");
-
-// /* Dimensions */
-// n = mxGetM(prhs[0]);
-// psin = mxGetN(prhs[3]);
-// pin = mxGetN(prhs[4]);
- n = gensys_ps->G0_dm->nrows;
- psin = gensys_ps->Psi_dm->ncols;
- pin = gensys_ps->Pi_dm->ncols;
- //
- eu = gensys_ps->eu_dv->v;
- //eu = mxGetPr(plhs[7]=mxCreateDoubleMatrix(2,1,mxREAL));
-
-// /* Check for consistency of input matrix dimensions */
-// if (mxGetM(prhs[0])!=n || mxGetM(prhs[1])!=n)
-// mexErrMsgTxt("g0 and g1 must be square matrices.\n");
-
-// if (mxGetM(prhs[3])!=n || mxGetM(prhs[4])!=n)
-// mexErrMsgTxt("psi and pi must have the same number of rows as g0.\n");
-
-// eu = mxGetPr(plhs[7]=mxCreateDoubleMatrix(2,1,mxREAL));
-
- //=== [a b q z]=qz(g0,g1);
-// a = mat2mkl(prhs[0],n,n);
-// b = mat2mkl(prhs[1],n,n);
-// q = mxCalloc(n*n,sizeof(MKL_Complex16));
-// z = mxCalloc(n*n,sizeof(MKL_Complex16));
- a = CreateComplexMatrix5RealMatrix(gensys_ps->G0_dm);
- b = CreateComplexMatrix5RealMatrix(gensys_ps->G1_dm);
- q = tzMalloc(nsquare=square(n), MKL_Complex16);
- z = tzMalloc(nsquare, MKL_Complex16);
- InitializeConstantMLK_Complex16(q, nsquare, 0.0);
- InitializeConstantMLK_Complex16(z, nsquare, 0.0);
-
-// fixdiv = nrhs != 6;
-// stake = fixdiv ? 1.01 : mxGetScalar(prhs[5]);
- fixdiv = (gensys_ps->div < 1.0);
- stake = fixdiv ? 1.01 : gensys_ps->div;
- nunstab = 0;
- zxz = 0;
-
- if (qz(a, b, q, z, n)) {
- printf("WARNING: QZ factorization failed.\n");
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
-
- nunstab /= 2;
-
- if (zxz) {
- printf("WARNING: Coincident zeros. Indeterminacy and/or nonexistence.\n");
- eu[0] = eu[1] = -2.0;
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
-
-// plhs[6] = mxCreateDoubleMatrix(n, 2, mxCOMPLEX); //For Gev_dzm.
-// copy_eigenvalues(plhs[6], a, b, n);
- copy_eigenvalues(gensys_ps->Gev_dzm, a, b);
-
- one.real = 1.0;
- one.imag = 0.0;
-
- minusone.real = -1.0;
- minusone.imag = 0.0;
-
- zero.real = 0.0;
- zero.imag = 0.0;
-
-// pi = mat2mkl(prhs[4],n,pin);
-// etawt = mxCalloc(nunstab*pin,sizeof(MKL_Complex16));
- pi = CreateComplexMatrix5RealMatrix(gensys_ps->Pi_dm);
- etawt = tzMalloc(tmpi=nunstab*pin, MKL_Complex16);
- InitializeConstantMLK_Complex16(etawt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, pin, n,
- &one, q+n*(n-nunstab), n, pi, n, &zero, etawt, nunstab);
- if (compute_svd(etawt, &ueta, &deta, &veta, nunstab, pin)) {
- //Memory is now allocated to ueta, deta, and veta.
- printf("WARNING: SVD failed.\n");
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(etawt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(etawt);
- md = nunstab<pin?nunstab:pin;
- bigev = md;
- for (i=0; i<md; i++)
- if (deta[i]<=REALSMALL) {
- bigev=i;
- break;
- }
-
-// psi = mat2mkl(prhs[3],n,psin);
-// zwt = mxCalloc(nunstab*psin,sizeof(MKL_Complex16));
- psi = CreateComplexMatrix5RealMatrix(gensys_ps->Psi_dm);
- zwt = tzMalloc(tmpi=nunstab*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(zwt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, psin, n,
- &one, q+n*(n-nunstab), n, psi, n, &zero, zwt, nunstab);
- if (compute_svd(zwt, &uz, &dz, &vz, nunstab, psin)) {
- //Memory is now allocated to uz, dz, and vz.
- printf("WARNING: SV decomposition failed.\n");
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(uz);
- tzDestroy(dz);
- tzDestroy(vz);
- tzDestroy(zwt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(vz);
- mds = nunstab<psin?nunstab:psin;
- bigevs = mds;
- for (i=0; i<mds; i++)
- if (dz[i]<=REALSMALL) {
- bigevs=i;
- break;
- }
- tzDestroy(dz);
-
-
- /* ueta = nunstab x bigev
- deta = bigev x 1
- veta = bigev x pin, ldveta = md
- uz = nunstab x bigevs
- dz = bigevs x 1
- vz = bigevs x psin, ldvz = mds */
-
- if (!bigevs) {
- exist = 1;
- existx = 1;
- } else {
- /* uz-ueta*ueta'*uz */
- MKL_Complex16 *tmp = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, nunstab, nunstab,
- bigev, &one, ueta, nunstab, ueta, nunstab, &zero, tmp, nunstab);
- cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, nunstab,
- bigevs, &minusone, tmp, nunstab, uz, nunstab, &one, uz, nunstab);
- tzDestroy(tmp);
- if (compute_norm(uz, &norm, nunstab, bigevs)) {
- //Memory is now allocated to norm.
- printf("WARNING: SVD failed.\n");
- tzDestroy(norm);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(uz);
- tzDestroy(zwt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- exist = *norm < REALSMALL*n;
- tzDestroy(norm);
- if (compute_normx(a, b, zwt, ueta, &normx, nunstab, psin, n, bigev)) {
- //If 0, memory is now allocated to normx; otherwise, normx is destroyed within the function compute_normx().
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(uz);
- tzDestroy(zwt);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- existx = *normx < REALSMALL*n;
- tzDestroy(normx);
- }
-
- tzDestroy(uz);
- tzDestroy(zwt);
-
-
-/* Note that existence and uniqueness are not just matters of comparing
- numbers of roots and numbers of endogenous errors. These counts are
- reported below because usually they point to the source of the problem. */
-
- etawt1 = tzMalloc(tmpi=(n-nunstab)*pin, MKL_Complex16);
- InitializeConstantMLK_Complex16(etawt1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, n-nunstab, pin, n,
- &one, q, n, pi, n, &zero, etawt1, n-nunstab);
- tzDestroy(pi);
- if (compute_svd(etawt1, &ueta1, &deta1, &veta1, n-nunstab, pin)) {
- //Memory is now allocated to ueta1, deta1, and veta1.
- printf("WARNING: SVD failed for compute_svd().\n");
- tzDestroy(ueta1);
- tzDestroy(deta1);
- tzDestroy(veta1);
- tzDestroy(etawt1);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(etawt1);
- md1 = n-nunstab<pin?n-nunstab:pin;
- bigev1 = md1;
- for (i=0; i<md1; i++)
- if (deta1[i]<=REALSMALL) {
- bigev1=i;
- break;
- }
- /* ueta1 = n-nunstab x bigev1
- deta1 = bigev1 x 1
- veta1 = bigev1 x pin, ldveta1 = md1 */
-
- if (existx || !nunstab) {
- //=== Solution exists.
- eu[0] = 1.0;
- } else {
- if (exist) {
- printf("WARNING: Solution exists for unforecastable z only\n");
- eu[0] = -1.0;
- } /* else
- mexPrintf("No solution. %d unstable roots. %d endog errors.\n",nunstab,bigev1); */
- /* mexPrintf("Generalized eigenvalues\n");
- mexCallMATLAB(0,NULL,1, &plhs[6], "disp"); */
- }
- if (!bigev1)
- unique = 1;
- else {
- /* veta1-veta1*veta*veta' */
- /* veta = bigev x pin, ldveta1 = md
- veta1 = bigev1 x pin, ldveta1 = md1 */
- MKL_Complex16 *tmp = tzMalloc(pin*pin, MKL_Complex16);
- MKL_Complex16 *veta1_copy = tzMalloc(pin*bigev1, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, pin*pin, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- InitializeConstantMLK_Complex16(veta1_copy, pin*bigev1, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, pin, pin,
- bigev, &one, veta, md, veta, md, &zero, tmp, pin);
- cblas_zdupe(bigev1,pin,veta1,md1,veta1_copy,bigev1);
- cblas_zhemm(CblasColMajor, CblasRight, CblasUpper, bigev1, pin,
- &minusone, tmp, pin, veta1_copy, bigev1, &one, veta1_copy, bigev1);
- tzDestroy(tmp);
- if (compute_norm(veta1_copy, &norm, bigev1, pin)) {
- //Memory is now allocated to norm.
- printf("WARNING: SVD failed.\n");
- tzDestroy(norm);
- tzDestroy(ueta1);
- tzDestroy(deta1);
- tzDestroy(veta1);
- tzDestroy(ueta);
- tzDestroy(deta);
- tzDestroy(veta);
- tzDestroy(veta1_copy);
- tzDestroy(a);
- tzDestroy(b);
- tzDestroy(q);
- tzDestroy(z);
- return;
- }
- tzDestroy(veta1_copy);
- unique = *norm < REALSMALL*n;
- tzDestroy(norm);
- }
- if (unique) {
- //=== Unique solution.
- eu[1] = 1.0;
- } else {
- eu[1] = 0.0;
- printf("WARNING: Indeterminacy. %d loose endog errors with eu being [%g, %g].\n",bigev1-bigev, eu[0], eu[1]);
- //printf("WARNING: Indeterminacy. %d loose endog errors with eu being [%g, %g].\n",bigev1-bigev, gensys_ps->eu_dv->v[0], gensys_ps->eu_dv->v[1]);
- //printf("WARNING: Indeterminacy. %d loose endog errors.\n",bigev1-bigev);
- /* mexPrintf("Generalized eigenvalues\n");
- mexCallMATLAB(0,NULL,1, &plhs[6], "disp"); */
- }
-
- cblas_zdscali(pin,deta,bigev,veta,md); /* veta' = deta\veta' */
- tzDestroy(deta);
- cblas_zdscale(pin,deta1,bigev1,veta1,md1); /* veta1' = deta1*veta1' */
- tzDestroy(deta1);
- etawt = tzMalloc(tmpi=nunstab*pin, MKL_Complex16); /* etawt = ueta*veta' */
- InitializeConstantMLK_Complex16(etawt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, nunstab, pin, bigev,
- &one, ueta, nunstab, veta, md, &zero, etawt, nunstab);
- tzDestroy(ueta);
- tzDestroy(veta);
- etawt1 = tzMalloc(tmpi=(n-nunstab)*pin, MKL_Complex16); /* etawt1 = ueta1*veta1' */
- InitializeConstantMLK_Complex16(etawt1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, pin, bigev1,
- &one, ueta1, n-nunstab, veta1, md1, &zero, etawt1, n-nunstab);
- tzDestroy(ueta1);
- tzDestroy(veta1);
- tmat = tzMalloc(tmpi=(n-nunstab)*nunstab, MKL_Complex16); /* tmat = etawt1*etawt' */
- InitializeConstantMLK_Complex16(tmat, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n-nunstab, nunstab, pin,
- &one, etawt1, n-nunstab, etawt, nunstab, &zero, tmat, n-nunstab);
- tzDestroy(etawt1);
- tzDestroy(etawt);
-
- g0 = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(g0, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdupe(n-nunstab, n, a, n, g0, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, nunstab, nunstab,
- &minusone, tmat, n-nunstab, a+(n-nunstab)*(n+1), n, &one, g0+(n-nunstab)*n, n);
- cblas_zcopy(nunstab, &one, 0, g0+(n-nunstab)*(n+1), n+1);
-
- g1 = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(g1, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdupe(n-nunstab, n, b, n, g1, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n-nunstab, nunstab, nunstab,
- &minusone, tmat, n-nunstab, b+(n-nunstab)*(n+1), n, &one, g1+(n-nunstab)*n, n);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, n, &one, g0, n, g1, n);
- dummy = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, n, n, &one, z, n, g1, n, &zero, dummy, n);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n, n, n, &one, dummy, n, z, n, &zero, g1, n);
- tzDestroy(dummy);
- //plhs[0] = mkl2mat_real(g1, n, n, n);
- ComplexMatrix2RealMatrix(gensys_ps->Theta_dm, g1); //Output.
- tzDestroy(g1);
-
- tmatq = tzMalloc(tmpi=n*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmatq, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zcopy(n*n, q, 1, tmatq, 1);
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, n, n-nunstab, nunstab,
- &minusone, tmatq+(n-nunstab)*n, n, tmat, n-nunstab, &one, tmatq, n);
- tzDestroy(tmat);
-
- ab = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(ab, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zdpsb(nunstab, nunstab, a+(n-nunstab)*(n+1), n, b+(n-nunstab)*(n+1), n, ab, nunstab);
- cblas_ztrsm(CblasColMajor, CblasRight, CblasUpper, CblasConjTrans, CblasNonUnit,
- n, nunstab, &one, ab, nunstab, tmatq+(n-nunstab)*n, n);
- tzDestroy(ab);
-
- //c = mat2mkl(prhs[2],n,1);
- c = CreateComplexMatrix5RealVector(gensys_ps->c0_dv);
- dummy = tzMalloc(gensys_ps->c0_dv->n, MKL_Complex16);
- //$$$$$$$ The following is Iskander's fatal code error. One cannot use c in the two different places; otherwise, it makes c be zero completely!
- // cblas_zgemv(CblasColMajor, CblasConjTrans, n, n, &one, tmatq, n, c, 1, &zero, c, 1);
- // cblas_ztrsv(CblasColMajor, CblasUpper, CblasNoTrans, CblasNonUnit, n, g0, n, c, 1);
- // cblas_zgemv(CblasColMajor, CblasNoTrans, n, n, &one, z, n, c, 1, &zero, c, 1);
- cblas_zgemv(CblasColMajor, CblasConjTrans, n, n, &one, tmatq, n, c, 1, &zero, dummy, 1);
- cblas_ztrsv(CblasColMajor, CblasUpper, CblasNoTrans, CblasNonUnit, n, g0, n, dummy, 1);
- cblas_zgemv(CblasColMajor, CblasNoTrans, n, n, &one, z, n, dummy, 1, &zero, c, 1);
- //plhs[1] = mkl2mat_real(c, n, n, 1);
- ComplexMatrix2RealVector(gensys_ps->c_dv, c); //Output.
- tzDestroy(c);
- tzDestroy(dummy);
-
- impact = tzMalloc(tmpi=n*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(impact, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, n-nunstab, psin, n,
- &one, tmatq, n, psi, n, &zero, impact, n);
- tzDestroy(tmatq);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, psin, &one, g0, n, impact, n);
- dummy = tzMalloc(tmpi=n*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, psin, n,
- &one, z, n, impact, n, &zero, dummy, n);
- tzDestroy(impact);
- //plhs[2] = mkl2mat_real(dummy, n, n, psin);
- ComplexMatrix2RealMatrix(gensys_ps->Impact_dm, dummy); //Output.
- tzDestroy(dummy);
-
- fmat = a+(n-nunstab)*(n+1);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, nunstab, &one, b+(n-nunstab)*(n+1), n, fmat, n);
- //plhs[3] = mkl2mat(fmat, n, nunstab, nunstab);
- gensys_ps->Fmat_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Fmat_dzm, fmat, n, nunstab, nunstab);
- tzDestroy(a);
-
- fwt = tzMalloc(tmpi=nunstab*psin, MKL_Complex16);
- InitializeConstantMLK_Complex16(fwt, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_ztrsm(CblasColMajor, CblasRight, CblasUpper, CblasConjTrans, CblasNonUnit,
- n, nunstab, &one, b+(n-nunstab)*(n+1), n, q+(n-nunstab)*n, n);
- tzDestroy(b);
- cblas_zgemm(CblasColMajor, CblasConjTrans, CblasNoTrans, nunstab, psin, n,
- &minusone, q+(n-nunstab)*n, n, psi, n, &zero, fwt, nunstab);
- tzDestroy(q);
- tzDestroy(psi);
- //plhs[4] = mkl2mat(fwt, nunstab, nunstab, psin);
- gensys_ps->Fwt_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Fwt_dzm, fwt, nunstab, nunstab, psin);
- tzDestroy(fwt);
-
- ywt = tzMalloc(tmpi=n*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(ywt, tmpi, 0.0);
- cblas_zcopy(nunstab, &one, 0, ywt+n-nunstab, n+1);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- n, nunstab, &one, g0, n, ywt, n);
- tzDestroy(g0);
- dummy = tzMalloc(tmpi=n*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(dummy, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, n, nunstab, n,
- &one, z, n, ywt, n, &zero, dummy, n);
- tzDestroy(z);
- tzDestroy(ywt);
- //plhs[5] = mkl2mat(dummy, n, n, nunstab);
- gensys_ps->Ywt_dzm = SubComplexMatrix2Zmatrix(gensys_ps->Ywt_dzm, dummy, n, n, nunstab);
- tzDestroy(dummy);
-}
-
-
-/******************************************************************************
- * function selctg orders the eigenvalues so that a selected cluster of *
- * eigenvalues appears in the leading diagonal blocks of the upper *
- * quasi-triangular matrix S and the upper triangular matrix T. *
- ******************************************************************************/
-
-static int selctg(MKL_Complex16 *alpha, MKL_Complex16 *beta)
-{
- double absA = sqrt(alpha->real*alpha->real+alpha->imag*alpha->imag),
- absB = fabs(beta->real);
- if (absA) {
- double divhat = absB/absA;
- //bug detected by Vasco Curdia and Daria Finocchiaro, 2/25/2004 CAS. A root of
- //exactly 1.01 and no root between 1 and 1.02, led to div being stuck at 1.01
- //and the 1.01 root being misclassified as stable. Changing < to <= below fixes this.
- if (fixdiv && 1+REALSMALL<divhat && divhat<=stake)
- stake = (1+divhat)/2;
- }
- if (absA<REALSMALL && absB<REALSMALL)
- zxz = 1;
- if (absB>stake*absA) {
- nunstab++;
- return(0);
- } else
- return(1);
-}
-
-/******************************************************************************
- * compute for a pair of N-by-N complex nonsymmetric matrices (A,B), *
- * the generalized eigenvalues, the generalized complex Schur form (S, T), *
- * and optionally left and/or right Schur vectors (VSL and VSR) *
- ******************************************************************************/
-
-static int qz(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *q, MKL_Complex16 *z, int n)
-{
-// unsigned char msg[101];
- int sdim, lwork = -1, info = 0;
- MKL_Complex16 *alpha = tzMalloc(n,MKL_Complex16),
- *beta = tzMalloc(n,MKL_Complex16),
- *work, work1;
- double *rwork = tzMalloc(8*n, double);
- int *bwork = tzMalloc(4*n, int);
-
- /* Query zgges on the value of lwork */
- zgges("V", "V", "S", &selctg, &n, a, &n, b, &n, &sdim, alpha, beta, q,
- &n, z, &n, &work1, &lwork, rwork, bwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to the Intel MKL function zgges() has an illegal value",-info);
- tzDestroy(bwork);
- tzDestroy(rwork);
- tzDestroy(alpha);
- tzDestroy(beta);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgges("V", "V", "S", &selctg, &n, a, &n, b, &n, &sdim, alpha, beta, q,
- &n, z, &n, work, &lwork, rwork, bwork, &info);
-
- tzDestroy(work);
- tzDestroy(bwork);
- tzDestroy(rwork);
- tzDestroy(alpha);
- tzDestroy(beta);
-
- if (info < 0) {
- printf("WARNING: Input %d to the Intel MKL function zgges() has an illegal value",-info);
- return(info);
- }
-
- if (info > 0)
- if (info < n)
- printf("WARNING: The QZ iteration failed. (A,B) are not in Schur form,\n"
- "but ALPHA(j) and BETA(j) should be correct for j=%d,...,N.",info+1);
- else {
- switch (info-n) {
- case 1:
- printf("WARNING: LAPACK problem: error return from ZGGBAL");
- break;
- case 2:
- printf("WARNING: LAPACK problem: error return from ZGEQRF");
- break;
- case 3:
- printf("WARNING: LAPACK problem: error return from ZUNMQR");
- break;
- case 4:
- printf("WARNING: LAPACK problem: error return from ZUNGQR");
- break;
- case 5:
- printf("WARNING: LAPACK problem: error return from ZGGHRD");
- break;
- case 6:
- printf("WARNING: LAPACK problem: error return from ZHGEQZ (other than failed iteration)");
- break;
- case 7:
- printf("WARNING: LAPACK problem: error return from ZGGBAK (computing VSL)");
- break;
- case 8:
- printf("WARNING: LAPACK problem: error return from ZGGBAK (computing VSR)");
- break;
- case 9:
- printf("WARNING: LAPACK problem: error return from ZLASCL (various places)");
- break;
- default:
- printf("WARNING: LAPACK problem: unknown error.");
- break;
- }
- }
- return(info);
-}
-
-/*
- * Convert MATLAB complex matrix to MKL complex storage.
- *
- * Z = mat2mkl(X,ldz,ndz)
- *
- * converts MATLAB's mxArray X to MKL_Complex16 Z(ldz,ndz).
- * The parameters ldz and ndz determine the storage allocated for Z,
- * while mxGetM(X) and mxGetN(X) determine the amount of data copied.
- */
-
-//MKL_Complex16* mat2mkl(const mxArray *X, int ldz, int ndz) {
-// MKL_Complex16 *Z, *zp;
-// int m, n, incz, cmplxflag;
-// register int i, j;
-// double *xr, *xi;
-
-// Z = mxCalloc(ldz*ndz, sizeof(MKL_Complex16));
-// xr = mxGetPr(X);
-// xi = mxGetPi(X);
-// m = mxGetM(X);
-// n = mxGetN(X);
-// zp = Z;
-// incz = ldz-m;
-// cmplxflag = (xi != NULL);
-// for (j = 0; j < n; j++) {
-// if (cmplxflag) {
-// for (i = 0; i < m; i++) {
-// zp->real = *xr++;
-// zp->imag = *xi++;
-// zp++;
-// }
-// } else {
-// for (i = 0; i < m; i++) {
-// zp->real = *xr++;
-// zp++;
-// }
-// }
-// zp += incz;
-// }
-// return(Z);
-//}
-
-
-/*
- * Convert MKL complex storage to MATLAB real and imaginary parts.
- *
- * X = mkl2mat(Z,ldz,m,n)
- *
- * copies MKL_Complex16 Z to X, producing a complex mxArray
- * with mxGetM(X) = m and mxGetN(X) = n.
- */
-
-//mxArray* mkl2mat(MKL_Complex16 *Z, int ldz, int m, int n) {
-// int i, j, incz;
-// double *xr, *xi;
-// MKL_Complex16 *zp;
-// mxArray *X;
-
-// X = mxCreateDoubleMatrix(m,n,mxCOMPLEX);
-// xr = mxGetPr(X);
-// xi = mxGetPi(X);
-// zp = Z;
-// incz = ldz-m;
-// for (j = 0; j < n; j++) {
-// for (i = 0; i < m; i++) {
-// *xr++ = zp->real;
-// *xi++ = zp->imag;
-// zp++;
-// }
-// zp += incz;
-// }
-// return(X);
-//}
-
-//plhs[3] = mkl2mat(fmat, n, nunstab, nunstab)
-
-/*
- * Convert MKL complex storage to MATLAB real matrix ignoring imaginary part.
- *
- * X = mkl2mat(Z,ldz,m,n)
- *
- * copies MKL_Complex16 Z to X, producing a real mxArray
- * with mxGetM(X) = m and mxGetN(X) = n.
- */
-
-//mxArray* mkl2mat_real(MKL_Complex16 *Z, int ldz, int m, int n) {
-// int i, j, incz;
-// double *xr;
-// MKL_Complex16 *zp;
-// mxArray *X;
-
-// X = mxCreateDoubleMatrix(m,n,mxREAL);
-// xr = mxGetPr(X);
-// zp = Z;
-// incz = ldz-m;
-// for (j = 0; j < n; j++) {
-// for (i = 0; i < m; i++) {
-// *xr++ = zp->real;
-// zp++;
-// }
-// zp += incz;
-// }
-// return(X);
-//}
-
-//void copy_eigenvalues(mxArray *gev, MKL_Complex16 *a, MKL_Complex16 *b, int n) {
-// double *gevr = mxGetPr(gev),
-// *gevi = mxGetPi(gev);
-// int i;
-
-// for (i=0; i<n; i++, gevr++, gevi++, a+=n+1) {
-// *gevr = a->real;
-// *gevi = a->imag;
-// }
-
-// for (i=0; i<n; i++, gevr++, gevi++, b+=n+1) {
-// *gevr = b->real;
-// *gevi = b->imag;
-// }
-//}
-
-static void copy_eigenvalues(TSdzmatrix *Gev_dzm, MKL_Complex16 *a, MKL_Complex16 *b)
-{
- int n = Gev_dzm->real->nrows;
- double *gevr = Gev_dzm->real->M,
- *gevi = Gev_dzm->imag->M;
- int i;
-
- for (i=0; i<n; i++, gevr++, gevi++, a+=n+1) {
- *gevr = a->real;
- *gevi = a->imag;
- }
-
- for (i=0; i<n; i++, gevr++, gevi++, b+=n+1) {
- *gevr = b->real;
- *gevi = b->imag;
- }
-}
-
-
-static int compute_svd(MKL_Complex16 *x, MKL_Complex16 **u, double **d, MKL_Complex16 **v, int m, int n)
-{
- //$$$Memory allocated to u, d, and v will be destroyed outside this function.$$$
- int tmpi;
- int md = m<n?m:n, lwork = -1, info = 0;
- MKL_Complex16 *a, *work, work1;
- double *rwork = tzMalloc(5*md>1?5*md:1, double);
-
- a = tzMalloc(m*n, MKL_Complex16);
- cblas_zcopy(m*n, x, 1, a, 1);
-
- *u = tzMalloc(tmpi=m*md,MKL_Complex16);
- InitializeConstantMLK_Complex16(*u, tmpi, 0.0);
- *v = tzMalloc(tmpi=md*n, MKL_Complex16);
- InitializeConstantMLK_Complex16(*v, tmpi, 0.0);
- *d = tzMalloc(md, double);
- InitializeConstantDouble(*d, md, 0.0);
-
- /* Query zgges on the value of lwork */
- zgesvd("S", "S", &m, &n, a, &m, *d, *u, &m, *v, &md, &work1, &lwork, rwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
- tzDestroy(rwork);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgesvd("S", "S", &m, &n, a, &m, *d, *u, &m, *v, &md, work, &lwork, rwork, &info);
-
- tzDestroy(work);
- tzDestroy(rwork);
- tzDestroy(a);
-
- if (info < 0)
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
-
- if (info > 0)
- printf("WARNING: ZBDSQR did not converge.\n%d superdiagonals of an intermediate "
- "bidiagonal form B did not converge to zero.",info);
- return(info);
-}
-
-static int compute_norm(MKL_Complex16 *a, double **d, int m, int n)
-{
- //Memory will be allocated to d, which will be destroyed outside this function.
- int md = m<n?m:n, lwork = -1, info = 0;
- MKL_Complex16 *work = NULL, work1;
- double *rwork = tzMalloc(5*md>1?5*md:1, double);
-
- *d = tzMalloc(md, double);
-
- /* Query zgges on the value of lwork */
- zgesvd("N", "N", &m, &n, a, &m, *d, NULL, &m, NULL, &md, &work1, &lwork, rwork, &info);
-
- if (info < 0) {
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
- tzDestroy(rwork);
- return(info);
- }
-
- lwork = (int)(work1.real);
- work = tzMalloc(lwork, MKL_Complex16);
- zgesvd("N", "N", &m, &n, a, &m, *d, NULL, &m, NULL, &md, work, &lwork, rwork, &info);
-
- tzDestroy(work);
- tzDestroy(rwork);
-
- if (info < 0)
- printf("WARNING: Input %d to zgesvd had an illegal value",-info);
-
- if (info > 0)
- printf("WARNING: ZBDSQR() in Intel MKL did not converge.\n%d superdiagonals of an intermediate "
- "bidiagonal form B did not converge to zero.",info);
-
- return(info);
-}
-
-static int compute_normx(MKL_Complex16 *a, MKL_Complex16 *b, MKL_Complex16 *zwt, MKL_Complex16 *ueta, double **normx, int nunstab, int psin, int n, int bigev)
-{
- //Memory is allocated to normx, which will be freed outside this function.
- int tmpi;
- int info = 0, i, bigevs;
- //
- MKL_Complex16 *M = NULL, *zwtx = NULL, *ux = NULL, *vx = NULL, *tmp = NULL;
- double *dx = NULL;
-
-
- a += (n+1)*(n-nunstab);
- b += (n+1)*(n-nunstab);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, psin, &one, b, n, zwt, nunstab);
- M = tzMalloc(nunstab*nunstab, MKL_Complex16);
- cblas_zdupe(nunstab, nunstab, a, n, M, nunstab);
- cblas_ztrsm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit,
- nunstab, nunstab, &one, b, n, M, nunstab);
-
- zwtx = tzMalloc(nunstab*nunstab*psin, MKL_Complex16);
- cblas_zcopy(nunstab*psin, zwt, 1, zwtx, 1);
- for (i=1; i<nunstab; i++) {
- cblas_ztrmm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, nunstab, psin*i, &one, M, nunstab, zwtx, nunstab);
- cblas_zcopy(nunstab*psin, zwt, 1, zwtx+nunstab*psin*i, 1);
- }
- tzDestroy(M);
- cblas_ztrmm(CblasColMajor, CblasLeft, CblasUpper, CblasNoTrans, CblasNonUnit, nunstab, nunstab*psin, &one, b, n, zwtx, nunstab);
- info = compute_svd(zwtx, &ux, &dx, &vx, nunstab, nunstab*psin); //Memory is allocated to ux, dx, and vx.
- tzDestroy(vx);
- tzDestroy(zwtx);
- if (info) {
- printf("WARNING: SVD failed.\n");
- tzDestroy(ux);
- tzDestroy(dx);
- return(info);
- }
- bigevs = nunstab;
- for (i=0; i<nunstab; i++)
- if (dx[i]<=REALSMALL) {
- bigevs = i;
- break;
- }
- tzDestroy(dx);
- /* ux-ueta*ueta'*ux */
- tmp = tzMalloc(tmpi=nunstab*nunstab, MKL_Complex16);
- InitializeConstantMLK_Complex16(tmp, tmpi, 0.0); //Must be initialized to 0.0 in order to have legal values of this pointer.
- cblas_zgemm(CblasColMajor, CblasNoTrans, CblasConjTrans, nunstab, nunstab,
- bigev, &one, ueta, nunstab, ueta, nunstab, &zero, tmp, nunstab);
- cblas_zhemm(CblasColMajor, CblasLeft, CblasUpper, nunstab,
- bigevs, &minusone, tmp, nunstab, ux, nunstab, &one, ux, nunstab);
- tzDestroy(tmp);
- info = compute_norm(ux, normx, nunstab, bigevs); //Memory is allocated to normx.
- if (info) {
- printf("WARNING: SVD failed.\n");
- tzDestroy(normx);
- tzDestroy(ux);
- return(info);
- }
- tzDestroy(ux);
- return(info);
-}
-
-static void cblas_zdupe(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<m; i++, a++, b++)
- cblas_zcopy(n, a, lda, b, ldb);
-}
-
-static void cblas_zdscali(int n, double *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<lda; i++, a++, b++)
- cblas_zdscal(n, 1.0/(*a), b, ldb);
-}
-
-static void cblas_zdscale(int n, double *a, int lda, MKL_Complex16 *b, int ldb)
-{
- int i;
- for (i=0; i<lda; i++, a++, b++)
- cblas_zdscal(n, *a, b, ldb);
-}
-
-static void cblas_zdpsb(int m, int n, MKL_Complex16 *a, int lda, MKL_Complex16 *b, int ldb, MKL_Complex16 *c, int ldc)
-{
- int i;
- cblas_zdupe(m, n, a, lda, c, ldc);
- for (i=0; i<m; i++, b++, c++)
- cblas_zaxpy(n, &minusone, b, ldb, c, ldc);
-}
-
-
-static MKL_Complex16* CreateComplexMatrix5RealMatrix(TSdmatrix *X_dm)
-{
- int mn, k;
- double *M;
- //
- MKL_Complex16 *Z = NULL;
-
- if (!X_dm) fn_DisplayError("CreateComplexMatrix5RealMatrix(): Input matrix X_dm must be allocated memory");
- M = X_dm->M;
-
- Z = tzMalloc(mn=X_dm->nrows*X_dm->ncols, MKL_Complex16);
- for (k=mn-1; k>=0; k--) {
- Z[k].real = M[k];
- Z[k].imag = 0.0;
- }
- return(Z);
-}
-
-static MKL_Complex16* CreateComplexMatrix5RealVector(TSdvector *x_dv)
-{
- int n, k;
- double *v;
- //
- MKL_Complex16 *Z = NULL;
-
- if (!x_dv) fn_DisplayError("CreateComplexMatrix5RealVector(): Input vector x_dv must be allocated memory");
- v = x_dv->v;
-
- Z = tzMalloc(n=x_dv->n, MKL_Complex16);
- for (k=n-1; k>=0; k--) {
- Z[k].real = v[k];
- Z[k].imag = 0.0;
- }
- return(Z);
-}
-
-
-static void ComplexMatrix2RealMatrix(TSdmatrix *Y_dm, MKL_Complex16 *Z)
-{
- int _k;
- double *M;
-
- if (!Y_dm) fn_DisplayError("ComplexMatrix2RealMatrix(): Output matrix Y_dm must be allocated memory");
- M = Y_dm->M;
-
- for (_k=Y_dm->nrows*Y_dm->ncols-1; _k>=0; _k--) M[_k] = Z[_k].real;
- Y_dm->flag = M_GE;
-}
-
-
-static void ComplexMatrix2RealVector(TSdvector *y_dv, MKL_Complex16 *Z)
-{
- int _k;
- double *v;
-
- if (!y_dv) fn_DisplayError("ComplexMatrix2RealVector(): Output matrix y_dv must be allocated memory");
- v = y_dv->v;
-
- for (_k=y_dv->n-1; _k>=0; _k--) v[_k] = Z[_k].real;
- y_dv->flag = V_DEF;
-}
-
-
-static TSdzmatrix *SubComplexMatrix2Zmatrix(TSdzmatrix *X_dzm, MKL_Complex16 *Z, const int nrowsforZ, const int _m, const int _n)
-{
- //X_dzm is _m-by_n comlex types where nrowsforZ <= _m and Z is nrowsforZ-by-_n.
- int _i, _j, incz;
- double *Mreal, *Mimag;
- MKL_Complex16 *zp;
- //
- TSdzmatrix *Y_dzm = NULL;
-
- if (!X_dzm || X_dzm->real->nrows != _m || X_dzm->real->ncols != _n) {
- DestroyMatrix_dz(X_dzm); //Destroys Y_dzm if already allocated memory to accommodate a possbible change of its dimension.
- Y_dzm = CreateMatrix_dz(_m, _n);
- }
- else Y_dzm = X_dzm;
-
- Mreal = Y_dzm->real->M;
- Mimag = Y_dzm->imag->M;
- zp = Z;
- if ((incz=nrowsforZ-_m)<0) fn_DisplayError("SubComplexMatrix2ZMatrix(): Number of rows for the input complex matrix Z must be greater that of the output Z matrix Y_dzm");
-
- for (_j=0; _j<_n; _j++) {
- for (_i=0; _i<_m; _i++) {
- *Mreal++ = zp->real;
- *Mimag++ = zp->imag;
- zp++;
- }
- zp += incz;
- }
- return (Y_dzm);
-}
-
-
-static void InitializeConstantMLK_Complex16(MKL_Complex16 *x_clx, const int _n, const double c)
-{
- int _i;
-
- for (_i=_n-1; _i>=0; _i--)
- x_clx[_i].real = x_clx[_i].imag = c;
-}
-
-static void InitializeConstantDouble(double *x_p, const int _n, const double c)
-{
- int _i;
-
- for (_i=_n-1; _i>=0; _i--) x_p[_i] = c;
-}
diff --git a/CFiles/gensysOldVersionWorks.h b/CFiles/gensysOldVersionWorks.h
deleted file mode 100755
index 9c6b23a98b82eb0a53168e6df96e500ff4db0044..0000000000000000000000000000000000000000
--- a/CFiles/gensysOldVersionWorks.h
+++ /dev/null
@@ -1,65 +0,0 @@
-/*******************************************************************
- * [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
- *
- * System given as
- * g0*y(t)=g1*y(t-1)+c+psi*z(t)+pi*eta(t),
- * with z an exogenous variable process and eta being endogenously determined
- * one-step-ahead expectational errors. Returned system is
- * y(t)=G1*y(t-1)+C+impact*z(t)+ywt*inv(I-fmat*inv(L))*fwt*z(t+1) .
- * If z(t) is i.i.d., the last term drops out.
- * If div or stake is omitted from argument list, a div>1 or stake>1 is calculated.
- * eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
- * existence only with not-serially correlated z(t); eu=[-2,-2] for coincident zeros.
- *
- * g0, g1: n-by-n matrices.
- * c: n-by-1 constant terms.
- * z(t): m-by-1 vector of exogenous residuals where m < n.
- * psi: n-by-m matrix.
- * eta(t): h-by-1 vector of expectational (endogenous) errors.
- * pi: n-by-h matrix.
- * div: a real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- *-------
- * G1 or Theta_dm: n-by-n matrices.
- * C: n-by-1 vector of constant terms.
- * impact: n-by-m matrix.
- * gev: n-by-2 z vector of stacked generalized eigenvalues where gev(;,2) ./ gev(:,1) = eig(g0, g1).
- * ywt: n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- * fmat: nunstab-by-nunstab z matrix where nunstab is the number of non-stable roots.
- * fwt: nunstab-by-m z matrix.
-********************************************************************/
-
-#ifndef __GENSYS_H__
- #define __GENSYS_H__
-
- #include "tzmatlab.h"
-
- #define REALSMALL 1e-7
-
- typedef struct TSgensys_tag {
- //=== Output arguments.
- TSdmatrix *Theta_dm; //n-by-n.
- TSdvector *c_dv; //n-by-1.
- TSdmatrix *Impact_dm; //n-by-m.
- TSdzmatrix *Fmat_dzm; //nunstab-by-nunstab z matrix. Initialized to NULL and will be dynamically allocated whenever gensys() is called.
- TSdzmatrix *Fwt_dzm; //nunstab-by-m z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- TSdzmatrix *Ywt_dzm; //n-by-nunstab z matrix of possible complex numbers. Initialized to NULL and dynamically allocated.
- TSdzmatrix *Gev_dzm; //n-by-2 z matrix of possible complex numbers.
- TSdvector *eu_dv; //2-by-1.
- //=== Function itself.
- void (*gensys)(struct TSgensys_tag *, void *);
- //=== Input arguments, which are all intialized to 0.0 and whose flags are set to M_GE.
- TSdmatrix *G0_dm; //n-by-n.
- TSdmatrix *G1_dm; //n-by-n.
- TSdvector *c0_dv; //n-by-1.
- TSdmatrix *Psi_dm; //n-by-m.
- TSdmatrix *Pi_dm; //n-by-k whtere k is the number of expectational errors.
- double div; //A real number dividing stable and unstable roots.. If < 1.0, a div>1.0 is calculated mechanically.
- } TSgensys;
- //
- typedef void TFlinratexp(struct TSgensys_tag *, void *); //For linear rational expectations models.
-
- struct TSgensys_tag *CreateTSgensys(TFlinratexp *func, const int _n, const int _m, const int _k, const double div);
- struct TSgensys_tag *DestroyTSgensys(struct TSgensys_tag *gensys_ps);
- void gensys_sims(struct TSgensys_tag *gensys_ps, void *dummy_ps);
-#endif
-
diff --git a/CFiles/gradcd_switch.c b/CFiles/gradcd_switch.c
deleted file mode 100755
index 95c417940435d8ba6736e23b20306e98ef69bdc9..0000000000000000000000000000000000000000
--- a/CFiles/gradcd_switch.c
+++ /dev/null
@@ -1,202 +0,0 @@
-
-#include "gradcd_switch.h"
-//#include "dw_array.h"
-//#include "m_array.h"
-//#include "dw_error.h"
-//#include "dw_rand.h"
-
-#include <math.h>
-#include <string.h>
-
-
-/*
- Assumes:
- model is a pointer to a properly initialized TStateModel.
-*/
-double logTimetLHgivenparameters(TStateModel *model, TSdvector *x_dv, int nmodelpars, int endt_base1)
-{
- //void *model is TStateModel *model
- //endt_base1: base-1 t.
- int s, t;
- double logvalue;
- PRECISION scale, u;
- TMarkovStateVariable *sv=model->sv;
-
- //--- Added by Zha.
- if (endt_base1 < 1 || endt_base1 > sv->nobs)
- {
- printf("gradcd_switch.c/logLHgivenparameters_time_t(): Input endt_base1 must be between 1 and T.\n");
- exit(EXIT_FAILURE);
- }
- if (nmodelpars > x_dv->n)
- {
- printf("gradcd_switch.c/logLHgivenparameters_time_t(): Number of model parameters, nmodelpars, must be <= x_dv->n.\n");
- exit(EXIT_FAILURE);
- }
- ConvertFreeParametersToQ(model, x_dv->v + nmodelpars);
- ConvertFreeParametersToTheta(model,x_dv->v);
-
-
-
- //====== Initializes prior to forward recursion if necessary ======
- if (model->routines->pInitializeForwardRecursion)
- model->routines->pInitializeForwardRecursion(model);
-
- //====== Get ergodic distribution if necessary ======
- if (sv->UseErgodic)
- if (!Ergodic(model->V[0],sv->Q))
- {
- printf("ForwardRecursion(): Ergodic distribution does not exist.\n");
- exit(0);
- }
-
- //====== forward recursion ======
- for (t=1; t <= sv->nobs; t++)
- {
- //------ compute Z[t] ------
- ProductMV(model->Z[t],sv->Q,model->V[t-1]);
-
- //------ compute log conditional probabilities and scale ------
- scale=MINUS_INFINITY;
- for (s=sv->nstates-1; s >= 0; s--)
- {
- if (ElementV(model->Z[t],s) > 0.0)
- {
- u=LogConditionalLikelihood(s,t,model);
- scale=AddScaledLogs(1.0,scale,ElementV(model->Z[t],s),u);
- ElementV(model->V[t],s)=log(ElementV(model->Z[t],s)) + u;
- }
- else
- ElementV(model->V[t],s)=MINUS_INFINITY;
- }
-
- //------ update L ------
- model->L+=scale;
-
- //--- Added by Zha.
- if (t==endt_base1) logvalue = scale;
-
- //------ scale V[t] ------
- for (s=sv->nstates-1; s >= 0; s--)
- if (ElementV(model->V[t],s) != MINUS_INFINITY)
- ElementV(model->V[t],s)=exp(ElementV(model->V[t],s) - scale);
- else
- ElementV(model->V[t],s)=0.0;
- }
-
- model->ValidForwardRecursion=1;
-
- return (logvalue);
-}
-
-
-
-
-//-------------------------------
-//Modified from fn_gradcd() in cstz.c for the conjugate gradient method I or II
-//-------------------------------
-#define STPS 1.0e-04 // 6.0554544523933391e-6 step size = pow(DBL_EPSILON,1.0/3)
-#define GRADMANUAL 0.0e+01 //Arbitrarily (manually) set gradient.
-void gradcd_switch(TSdvector *g_dv, TStateModel *model, TSdvector *x_dv, int nmodelpars, int endt_base1, double (*fcn)(TStateModel *model, TSdvector *x_dv, int nmodelpars, int endt_base1), TSdvector *grdh_dv, double f0)
-{
- //Outputs:
- // g: the gradient n-by-1 g (no need to be initialized).
- //Inputs:
- // x: the vector point at which the gradient is evaluated. No change in the end although will be added or substracted by dh during the function (but in the end the original value will be put back).
- // n: the dimension of g or x.
- // fcn(): the function for which the gradient is evaluated
- // grdh: step size. If NULL, then dh is set automatically; otherwise, grdh is taken as a step size, often set as 1.0e-004.
- // f0: the value of (*fcn)(x). NOT used in this function except dealing with the boundary (NEARINFINITY) for the
- // minimization problem, but to be compatible with a genral function call where, say, gradfw_gen() and cubic
- // interpolation of central difference method will use f0.
-
- double *g, *x, *grdh;
- double dh, dhi, dh2i, fp, fm, tmp, *xp;
- int n, i;
-
- g = g_dv->v;
- x = x_dv->v;
- if (!grdh_dv) grdh = NULL;
- else grdh = grdh_dv->v;
- n = g_dv->n;
- if (n != x_dv->n)
- {
- printf("gradcd_switch.c/gradcd_switch(): Input vectors g_dv and x_dv must have the same length.\n");
- exit(EXIT_FAILURE);
- }
- if (grdh_dv && n != grdh_dv->n)
- {
- printf("gradcd_switch.c/gradcd_switch(): Input vector grdh_dv mush have the same length as x_dv.\n");
- exit(EXIT_FAILURE);
- }
-
- if (grdh) {
- //=== If f0 >= NEARINFINITY, we're in a bad region and so we assume it's flat in this bad region. This assumption may or may not work for a third-party optimimization routine.
- if (f0 >= NEARINFINITY)
- {
- for (i=n-1; i>=0; i--)
- g[i] = GRADMANUAL;
- return;; //Early exit.
- }
-
- dh2i = (dhi=1.0/(dh=*grdh))/2.0;
- for (i=0, xp=x; i<n; i++, xp++, g++) {
- tmp = *xp;
- *xp += dh;
- //The following statement is bad because dh does not get reset at the beginning of the loop and thus may get changed continually within the loop.
- // dh = *xp - tmp; // This increases the precision slightly.
- fp = fcn(model, x_dv, nmodelpars, endt_base1);
- //fp = fcn(n,x); // IMSL
- //fcn(n,x,&fp,NULL,NULL); /* NAG */
- *xp = tmp - dh;
- fm = fcn(model, x_dv, nmodelpars, endt_base1);
- //fm = fcn(n,x); // IMSL
- //fcn(n,x,&fm,NULL,NULL);
-
- //=== Checking the boundary condition for the minimization problem.
- if ((fp < NEARINFINITY) && (fm < NEARINFINITY)) *g = (fp-fm)*dh2i;
- else if (fp < NEARINFINITY) *g = (fp-f0)*dhi;
- else if (fm < NEARINFINITY) *g = (f0-fm)*dhi;
- else *g = GRADMANUAL;
-
- *xp = tmp; // Put the original value of x[i] back to x[i] so that the content x[i] is still unaltered.
- }
-
- }
- else {
- //=== If f0 >= NEARINFINITY, we're in a bad region and so we assume it's flat in this bad region. This assumption may or may not work for a third-party optimimization routine.
- if (f0 >= NEARINFINITY)
- {
- for (i=n-1; i>=0; i--)
- g[i] = GRADMANUAL;
- return;; //Early exit.
- }
-
- for (i=0, xp=x; i<n; i++, xp++, g++) {
- dh = fabs(*xp)<=1 ? STPS : STPS*(*xp);
- tmp = *xp;
- *xp += dh;
- dh = *xp - tmp; // This increases the precision slightly.
- fp = fcn(model, x_dv, nmodelpars, endt_base1);
- //fp = fcn(n,x); // IMSL
- //fcn(n,x,&fp,NULL,NULL); /* NAG */
- *xp = tmp - dh;
- fm = fcn(model, x_dv, nmodelpars, endt_base1);
- //fm = fcn(n,x); // IMSL
- //fcn(n,x,&fm,NULL,NULL);
-
- //=== Checking the boundary condition for the minimization problem.
- if ((fp < NEARINFINITY) && (fm < NEARINFINITY)) *g = (fp-fm)/(2.0*dh);
- else if (fp < NEARINFINITY) *g = (fp-f0)/dh;
- else if (fm < NEARINFINITY) *g = (f0-fm)/dh;
- else *g = GRADMANUAL;
-
- *xp = tmp; // Put the original value of x[i] back to x[i] so that the content x[i] is still unaltered.
- }
- }
-
- g_dv->flag = V_DEF;
-}
-#undef STPS
-#undef GRADMANUAL
-
diff --git a/CFiles/gradcd_switch.h b/CFiles/gradcd_switch.h
deleted file mode 100755
index e6b7097286968bea7d7b822e3090c12b084d9942..0000000000000000000000000000000000000000
--- a/CFiles/gradcd_switch.h
+++ /dev/null
@@ -1,6 +0,0 @@
-
-#include "switch.h"
-
-
-double logTimetLHgivenparameters(TStateModel *model, TSdvector *x_dv, int nmodelpars, int endt_base1);
-void gradcd_switch(TSdvector *g_dv, TStateModel *model, TSdvector *x_dv, int nmodelpars, int endt_base1, double (*fcn)(TStateModel *model, TSdvector *x_dv, int nmodelpars, int endt_base1), TSdvector *grdh_dv, double f0);
diff --git a/CFiles/kalman.c b/CFiles/kalman.c
old mode 100755
new mode 100644
index d9b4df9a68204ba7910c82c704588b878106f01e..dbeff14b363bd11f089876a8672873eb2d0903aa
--- a/CFiles/kalman.c
+++ b/CFiles/kalman.c
@@ -1,2908 +1,2909 @@
-/*===============================================================================================================
- * Check $$$ for important notes.
- * Check <<>> for updating DW's new switch code or questions for DW.
- *
- * kalcvf_urw(): the Kalman filter forward prediction specialized for only a univariate random walk (urw) process.
- *
- * State space model is defined as follows:
- * z(t+1) = z(t)+eta(t) (state or transition equation)
- * y(t) = x(t)'*z(t)+eps(t) (observation or measurement equation)
- * where for this function, eta and eps must be uncorrelated; y(t) must be 1-by-1. Note that
- * x(t): k-by-1;
- * z(t): k-by-1;
- * eps(t): 1-by-1 and ~ N(0, sigma^2);
- * eta(t): ~ N(0, V) where V is a k-by-k covariance matrix.
- *
- *
- * Written by Tao Zha, May 2004.
- * Revised, May 2008;
-=================================================================================================================*/
-
-/**
-//=== For debugging purpose.
-if (1)
-{
- double t_loglht;
-
- t_loglht = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, " %10.5f\n", t_loglht);
-
- fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
- fprintf(FPTR_DEBUG, "\n wP0_dv:\n");
- WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "\n Vt_dc->C[sti_v=%d]:\n", sti_v);
- WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
- fflush(FPTR_DEBUG);
-}
-/**/
-
-
-#include "kalman.h"
-
-
-static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
-static int Updatekalfilms_1stapp(int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv) //, int storeZ, int storeV)
-{
- int _i;
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
- //---
- TSkalcvfurw *kalcvfurw_ps = tzMalloc(1, TSkalcvfurw);
-
-
- kalcvfurw_ps->indx_tvsigmasq = tv;
- kalcvfurw_ps->fss = T;
- kalcvfurw_ps->kx = k;
-
- //===
- kalcvfurw_ps->V_dm = CreateMatrix_lf(k, k);
- kalcvfurw_ps->ylhtran_dv = CreateVector_lf(T);
- kalcvfurw_ps->Xrhtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->z10_dv = CreateVector_lf(k);
- kalcvfurw_ps->P10_dm = CreateMatrix_lf(k, k);
-
- kalcvfurw_ps->zupdate_dv = CreateVector_lf(k);
- kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->ylhtranpred_dv = CreateVector_lf(T);
- //
- rows_iv = CreateVector_int(T);
- cols_iv = CreateVector_int(T);
- for (_i=T-1; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // if (!storeZ) kalcvfurw_ps->Zpredtran_dm = (TSdmatrix *)NULL;
- // else kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- // if (!storeV) kalcvfurw_ps->Ppred_dc = (TSdcell *)NULL;
- // else {
- // rows_iv = CreateVector_int(T);
- // cols_iv = CreateVector_int(T);
- // for (_i=T; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- // kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // }
-
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
- return (kalcvfurw_ps);
-}
-
-TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps)
-{
- if (kalcvfurw_ps) {
- DestroyMatrix_lf(kalcvfurw_ps->V_dm);
- DestroyVector_lf(kalcvfurw_ps->ylhtran_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Xrhtran_dm);
- DestroyVector_lf(kalcvfurw_ps->z10_dv);
- DestroyMatrix_lf(kalcvfurw_ps->P10_dm);
-
- DestroyVector_lf(kalcvfurw_ps->zupdate_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Zpredtran_dm);
- DestroyCell_lf(kalcvfurw_ps->Ppred_dc);
- DestroyVector_lf(kalcvfurw_ps->ylhtranpred_dv);
-
- free(kalcvfurw_ps);
- return ((TSkalcvfurw *)NULL);
- }
- else return (kalcvfurw_ps);
-}
-
-
-void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps)
-{
- //See the notes of SWZ regarding the government's updating of the parameters in their Phillips-curve equation.
- //NOTE: make sure that the value of kalcvfurw_ps->sigmasq and other input values are given.
- int ti;
- double workd, workdenominv;
- //---
- int fss, kx;
- double sigmasq_fix = kalcvfurw_ps->sigmasq;
-// double sigmasq;
- TSdmatrix *V_dm;
- TSdmatrix *Zpredtran_dm;
- TSdcell *Ppred_dc;
- TSdvector *ylhtran_dv;
- TSdmatrix *Xrhtran_dm;
- //===
- TSdvector *workkxby1_dv = NULL; //kx-by-1.
-// TSdvector *work1kxby1_dv = NULL; //kx-by-1.
- TSdmatrix *workkxbykx_dm = NULL; //kx-by-kx symmetric and positive positive.
-// //===
-// TSdvector *zbefore_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vbefore_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
-// TSdvector *zafter_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vafter_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
- //******* WARNING: Some dangerous pointer movement to gain efficiency *******
-// double *yt_p;
-// double *Vbefore_p;
-// double *Vafter_p;
- TSdvector xt_sdv;
- TSdvector zbefore_sdv;
- //TSdmatrix Vbefore_sdm;
- TSdvector zafter_sdv;
- //TSdmatrix Vafter_sdm;
-
-
- if (!kalcvfurw_ps) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps must be created");
- if (!kalcvfurw_ps->V_dm || !kalcvfurw_ps->ylhtran_dv || !kalcvfurw_ps->Xrhtran_dm || !kalcvfurw_ps->z10_dv || !kalcvfurw_ps->P10_dm)
- fn_DisplayError(".../kalcvf_urw(): input arguments kalcvfurw_ps->V_dm, kalcvfurw_ps->ylhtran_dv, kalcvfurw_ps->Xrhtran_dm, kalcvfurw_ps->z10_dv, kalcvfurw_ps->P10_dm must be given legal values");
- if (!(kalcvfurw_ps->P10_dm->flag & (M_SU | M_SL))) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps->P10_dm must be symmetric");
- fss = kalcvfurw_ps->fss;
- kx = kalcvfurw_ps->kx;
- V_dm = kalcvfurw_ps->V_dm;
- Zpredtran_dm = kalcvfurw_ps->Zpredtran_dm;
- Ppred_dc = kalcvfurw_ps->Ppred_dc;
- ylhtran_dv = kalcvfurw_ps->ylhtran_dv;
- Xrhtran_dm = kalcvfurw_ps->Xrhtran_dm;
- //---
- xt_sdv.n = kx;
- xt_sdv.flag = V_DEF;
- zbefore_sdv.n = kx;
- zbefore_sdv.flag = V_DEF;
- zafter_sdv.n = kx;
- zafter_sdv.flag = V_DEF;
-
- //=== Memory allocation.
- workkxby1_dv = CreateVector_lf(kx);
- workkxbykx_dm = CreateMatrix_lf(kx, kx);
-
-
- //------- The first period (ti=0). -------
- zbefore_sdv.v = kalcvfurw_ps->z10_dv->v;
- zafter_sdv.v = Zpredtran_dm->M;
- xt_sdv.v = Xrhtran_dm->M;
- //---
-
- workd = ylhtran_dv->v[0] - (kalcvfurw_ps->ylhtranpred_dv->v[0]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, kalcvfurw_ps->P10_dm, &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
-
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(kalcvfurw_ps->z10_dv->v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t];
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
-
-
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[0], kalcvfurw_ps->P10_dm, workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[0]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
-
- //------- The rest of the periods (ti=1:T-1). -------
- for (ti=1; ti<fss; ti++) {
- //NOTE: ti=0 has been taken care of outside of this loop.
- zbefore_sdv.v = Zpredtran_dm->M + (ti-1)*kx;
- zafter_sdv.v = Zpredtran_dm->M + ti*kx;
- xt_sdv.v = Xrhtran_dm->M + ti*kx;
- //---
- workd = ylhtran_dv->v[ti] - (kalcvfurw_ps->ylhtranpred_dv->v[ti]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, Ppred_dc->C[ti-1], &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(zbefore_sdv.v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[ti], Ppred_dc->C[ti-1], workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[ti]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
- }
- CopyVector0(kalcvfurw_ps->zupdate_dv, &zafter_sdv);
- Zpredtran_dm->flag = M_GE;
- kalcvfurw_ps->ylhtranpred_dv->flag = V_DEF;
-
-// DestroyVector_lf(zbefore_dv);
-// DestroyMatrix_lf(Vbefore_dm);
-// DestroyVector_lf(zafter_dv);
-// DestroyMatrix_lf(Vafter_dm);
-
- DestroyVector_lf(workkxby1_dv);
-// DestroyVector_lf(work1kxby1_dv);
- DestroyMatrix_lf(workkxbykx_dm);
-}
-
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- General constant (known-time-varying) Kalman filter for DSGE models.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T)
-{
- int _i;
- //===
- TSivector *rows_iv = CreateVector_int(T);
- TSivector *cols_iv = CreateVector_int(T);
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfiltv_tag *kalfiltv_ps = tzMalloc(1, struct TSkalfiltv_tag);
-
-
- //--- Default value.
- kalfiltv_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfiltv_ps->ny = ny;
- kalfiltv_ps->nz = nz;
- kalfiltv_ps->T = T;
-
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfiltv_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfiltv_ps->at_dm = CreateMatrix_lf(ny, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = nz;
- }
- rows_iv->flag = cols_iv->flag = V_DEF;
- kalfiltv_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->bt_dm = CreateMatrix_lf(nz, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- kalfiltv_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->z0_dv = CreateVector_lf(nz);
- kalfiltv_ps->P0_dm = CreateMatrix_lf(nz, nz);
-
-
- //---
- kalfiltv_ps->zt_tm1_dm = CreateMatrix_lf(nz, T);
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Pt_tm1_dc = CreateCell_lf(rows_iv, cols_iv);
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfiltv_ps);
-
-}
-//---
-struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- if (kalfiltv_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfiltv_ps->yt_dm);
- DestroyMatrix_lf(kalfiltv_ps->at_dm);
- DestroyCell_lf(kalfiltv_ps->Ht_dc);
- DestroyCell_lf(kalfiltv_ps->Rt_dc);
- DestroyCell_lf(kalfiltv_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfiltv_ps->bt_dm);
- DestroyCell_lf(kalfiltv_ps->Ft_dc);
- DestroyCell_lf(kalfiltv_ps->Vt_dc);
- //---
- DestroyVector_lf(kalfiltv_ps->z0_dv);
- DestroyMatrix_lf(kalfiltv_ps->P0_dm);
- //---
- DestroyMatrix_lf(kalfiltv_ps->zt_tm1_dm);
- DestroyCell_lf(kalfiltv_ps->Pt_tm1_dc);
-
-
- //---
- tzDestroy(kalfiltv_ps); //Must be freed last!
-
- return ((struct TSkalfiltv_tag *)NULL);
- }
- else return (kalfiltv_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- New code: Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp2(int ny, int nz, int nu, int ne, int nst, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //=== Default value.
- kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
- kalfilmsinputs_1stapp_ps->ztm1_track = -1;
- kalfilmsinputs_1stapp_ps->dtm1_track = -1;
-
- //--- Other key assignments.
- kalfilmsinputs_1stapp_ps->ny = ny;
- kalfilmsinputs_1stapp_ps->nz = nz;
- kalfilmsinputs_1stapp_ps->nu = nu;
- kalfilmsinputs_1stapp_ps->ne = ne;
- kalfilmsinputs_1stapp_ps->nst = nst;
- kalfilmsinputs_1stapp_ps->T = T;
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- if (nu)
- {
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nu);
- kalfilmsinputs_1stapp_ps->Psiut_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- kalfilmsinputs_1stapp_ps->Psiut_dc = NULL;
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ne);
- kalfilmsinputs_1stapp_ps->Psiet_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- //--- For output arguments.
- rows_iv = CreateConstantVector_int(T+1, nz);
- cols_iv = CreateConstantVector_int(T+1, nst);
- kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_1stapp_ps);
-}
-//---
-struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp2(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- if (kalfilmsinputs_1stapp_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Psiut_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Psiet_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
- //---
- tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
- }
- else return (kalfilmsinputs_1stapp_ps);
-};
-
-struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //=== Default value.
- kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
- kalfilmsinputs_1stapp_ps->ztm1_track = -1;
- kalfilmsinputs_1stapp_ps->dtm1_track = -1;
-
- //--- Other key assignments.
- kalfilmsinputs_1stapp_ps->ny = ny;
- kalfilmsinputs_1stapp_ps->nz = nz;
- kalfilmsinputs_1stapp_ps->nst = nst;
- kalfilmsinputs_1stapp_ps->T = T;
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- //--- For output arguments.
- rows_iv = CreateConstantVector_int(T+1, nz);
- cols_iv = CreateConstantVector_int(T+1, nst);
- kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_1stapp_ps);
-}
-//---
-struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- if (kalfilmsinputs_1stapp_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
- //---
- tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
- }
- else return (kalfilmsinputs_1stapp_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_tag *kalfilmsinputs_ps = tzMalloc(1, struct TSkalfilmsinputs_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //--- Default value.
- kalfilmsinputs_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfilmsinputs_ps->ny = ny;
- kalfilmsinputs_ps->nz = nz;
- kalfilmsinputs_ps->nRc = nRc;
- kalfilmsinputs_ps->nRstc = nRstc;
- kalfilmsinputs_ps->nRv = nRv;
- kalfilmsinputs_ps->indxIndRegimes = indxIndRegimes;
- kalfilmsinputs_ps->T = T;
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_ps->at_dm = CreateMatrix_lf(ny, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, ny);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, ny);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_ps->bt_dm = CreateMatrix_lf(nz, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, nz);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- if (indxIndRegimes)
- {
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRc*nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- //--- For output arguments.
- if (indxIndRegimes)
- {
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRc*nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_ps);
-
-}
-//---
-struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps)
-{
- if (kalfilmsinputs_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->z0_dm);
- DestroyCell_lf(kalfilmsinputs_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_ps->Pt_tm1_d4);
- //---
- tzDestroy(kalfilmsinputs_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_tag *)NULL);
- }
- else return (kalfilmsinputs_ps);
-};
-
-
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//-- Constant-parameters (known-time-varying) Kalman filter
-//-----------------------------------------------------
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //General constant (known-time-varying) Kalman filter for DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- int T = kalfiltv_ps->T;
- int Tp1 = T + 1;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- int indx_badlh = 0; //1: bad likelihood with, say, -infinity of the LH value.
- int tdata, ti;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv; //double loglh_tdata; //logdetDtdata.
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- //--- Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
-
-
-
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "Fatal error: tz_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[0], P0_dm);
-
- //====== See p.002 in LiuWZ. ======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++ )
- {
- //Base-0 timing.
- ti = tdata + 1; //Next period.
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (kalfiltv_ps->loglh = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh += -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //loglh += -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (ti<T)
- {
- //Updating only up to tdata=T-2. The values at ti=T or tdata=T-1 will not be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[ti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + ti*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[ti], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + ti*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[ti], Vt_dc->C[ti]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[ti], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[ti], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[ti], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[ti], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (kalfiltv_ps->loglh = loglh);
-}
-/**
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //This function is used to test tz_logTimetCondLH_kalfiltv().
- int T = kalfiltv_ps->T;
- int tdata;
- double loglh;
-
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++) loglh += tz_logTimetCondLH_kalfiltv(0, tdata+1, kalfiltv_ps);
-
- return (loglh);
-}
-/**/
-//-----------------------------------------------------
-//-- Updating Kalman filter at time t for constant-parameters (or known-time-varying) Kalman filter.
-//-----------------------------------------------------
-double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //st: base-0 grand regime at time t, which is just a dummy for this constant-parameter function in order to use
- // Waggoner's automatic functions.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
- //
- //log LH at time t for constant (known-time-varying) Kalman-filter DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood at time t.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // April 2008, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //--- Output arguments.
- double loglh_timet; //log likelihood at time t.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
- //--- Input arguments.
- int tdata, tp1;
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- int T = kalfiltv_ps->T;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
-
-
- tdata = (tp1=inpt) - 1; //Base-0 time.
-
- //======= Initial condition. =======
- if (tdata==0)
- {
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_logTimetCondLH_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(FPTR_DEBUG, "Fatal error: tz_logTimetCondLH_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and thus the initial conditions must be supplied by, say, input arguments");
- fflush(FPTR_DEBUG);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[tdata], P0_dm);
- }
-
-
- //======= Liklihood at time t (see p.002 in LiuWZ). =======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //======= Updating for the next period. =======
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (tp1<T)
- {
- //Updating only up to tdata=T-2, because the values at tp1=T or tdata=T-1 will NOT be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[tp1], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + tp1*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[tp1], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + tp1*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[tp1], Vt_dc->C[tp1]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[tp1], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[tp1], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[tp1], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[tp1], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (loglh_timet);
-}
-
-
-
-
-//-----------------------------------------------------
-//- WARNING: bedore using this function, make sure to call the following functions
-// Only once in creating lwzmodel_ps: Refresh_kalfilms_*(lwzmodel_ps);
-// Everytime when parameters are changed: RefreshEverything(); RefreRunningGensys_allcases(lwzmodel_ps) in particular.
-//-----------------------------------------------------
-double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //st: base-0 grand regime -- deals with the cross-section values at time t.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
- //-- Output arguments
- double loglh_timet;
- //--- Input arguments
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- Local variables
- int tbase0;
- double logdet_Dtdata;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- TSdvector etdata_sdv;
- //=== Work arguments.
- TSdvector *wny_dv = CreateVector_lf(ny);
-
-
-
- //--- Critical checking.
- if (inpt > kalfilmsinputs_1stapp_ps->T)
- fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): The time exceeds the\n"
- " data sample size allocated the structure TSkalfilmsinputs_1stapp_tag");
-
- //--- The following is for safe guard. InitializeKalman_z10_P10() should be called in, say, RefreshEverything().
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- if (!InitializeKalman_z10_P10(kalfilmsinputs_1stapp_ps, (TSdmatrix *)NULL, (TSdcell *)NULL))
- fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): the system is non-stationary when calling"
- " InitializeKalman_z10_P10(). Please call this function in RefreshEverthing() and"
- " set the likehood to be -infty for early exit");
-
- tbase0=inpt-1;
-
- //------------------- The order matters. Updatekalfilms_1stapp() must be called before Update_et_Dt_1stapp(). -----------------
- //--- $$$ Critical updating where we MUSt have inpt-1. If inpt, Updatekalfilms_1stapp() will call this function again
- //--- $$$ because DW function ProbabilityStateConditionalCurrent() need to access this function at time inpt,
- //--- $$$ which has not computed before Updatekalfilms_1stapp(). Thus, we'll have an infinite loop.
- Updatekalfilms_1stapp(tbase0, kalfilmsinputs_1stapp_ps, smodel_ps);
-// //--- $$$ Critical updating.
-// Update_et_Dt_1stapp(tbase0, kalfilmsinputs_1stapp_ps);
-// //This function will give Dtdata_d4->F[tbase0], etdata_dc->C[tbase0], and PHtran_tdata_d4->F[tbase0].
-
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- //--- Forming the log conditional likelihood at t.
- etdata_sdv.n = ny;
- etdata_sdv.v = etdata_dc->C[tbase0]->M + ny*st;
- etdata_sdv.flag = V_DEF;
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_d4->F[tbase0]->C[st]))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, &etdata_sdv, '/', Dtdata_d4->F[tbase0]->C[st]);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, &etdata_sdv);
- //Done with all w*_dv.
-
- //===
- DestroyVector_lf(wny_dv);
-
- return (loglh_timet);
-}
-//======================================================
-//= Computing z_{1|0} and P_{1|0} for each new parameter values.
-//======================================================
-int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc)
-{
- //See p.001 and p.004 in LWZ Model II.
- //Outputs:
- // return 1: success in initializing; 0: initializing fails, so the likelihood must be set to -infty outside this function.
- // ztm1_track to track the time up to which Kalman filter have been updated.
- // z0_dm, zt_tm1_dc->C[0]
- // P0_dc, Pt_tm1_d4->F[0]
-
- //--- Output arguments
- TSdmatrix *z0_0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdmatrix *z0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdcell *P0_dc = kalfilmsinputs_1stapp_ps->P0_dc; //nz-by-nz-by-nst.
- //+ Used to get zt_tm1_dc->C[0] and Pt_tm1_d4->F[0] only.
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Input arguments
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //--- Local variables
- int sti;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv, z0_0_sdv, bt_sdv;
- TSdvector yt_sdv, at_sdv;
- //--- For the initial conditions: eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- //===
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
-
-
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- {
- z0_sdv.n = z0_0_sdv.n = bt_sdv.n = nz;
- z0_sdv.flag = z0_0_sdv.flag = bt_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (!kalfilmsinputs_1stapp_ps->indxIni)
- {
- z0_0_dm->flag = z0_dm->flag = M_GE;
- for (sti=nst-1; sti>=0; sti--)
- {
- if (kalfilmsinputs_1stapp_ps->DiffuseScale) //Diffuse initial conditions are used.
- {
- //--- Diffuse condition for z0_dv.
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
- bt_sdv.v = bt_dm->M + bt_sdv.n*sti;
- InitializeConstantVector_lf(&z0_0_sdv, 0.0);
- MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
- VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
- //--- Diffuse condition for P0_dm.
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, kalfilmsinputs_1stapp_ps->DiffuseScale); //To be used for DiffuseScale*I(nz)
- CopyMatrix0(P0_dc->C[sti], Wnzbynz_dm);
- //Done with W*_dm.
- }
- else //Unconditional moments for initial conditions are used.
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti]);
- if (errflag) fn_DisplayError("kalman.c/InitializeKalman_z10_P10(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0-SQRTEPSILON)) //(1.0+EPSILON))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,sti))\b(:,sti);
- z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti]);
- CopySubmatrix2vector(&z0_0_sdv, 0, bt_dm, 0, sti, bt_dm->nrows);
- bdivA_rgens(&z0_0_sdv, &z0_0_sdv, '\\', Wnzbynz_dm);
- //- Under the assumption s_0 = s_1 (this is a short-cut).
- MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
- VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,sti),F(:,:,sti)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti], Ft_dc->C[sti]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- if (0) //0: no printing.
- {
- #if defined (USE_DEBUG_FILE)
- fprintf(FPTR_DEBUG, "\n-------WARNING: ----------\n");
- fprintf(FPTR_DEBUG, "\nIn grand regime sti=%d\n", sti);
- fprintf(FPTR_DEBUG, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #else
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn grand regime sti=%d\n", sti);
- fprintf(stdout, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #endif
- }
- //=== See p.000.3 in LWZ Model II.
- //=== Do NOT use the following option. It turns out that this will often generate explosive conditional liklihood
- //=== at the end of the sample, because Pt_tm1 shrinks to zero overtime due to the sigularity of
- //=== the initila condition P_{1|0}.
- //--- Letting z0_dv = 0.0
- // z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- // InitializeConstantVector_lf(&z0_sdv, 0.0);
- // //--- Letting P0_dm = V
- // CopyMatrix0(P0_dc->C[sti], Vt_dc->C[sti]);
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (0); //Early exit with kalfilmsinputs_1stapp_ps->ztm1_track continues to be -1.
- }
- }
- }
- }
- else
- {
- if (!z10_dm) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition z_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyMatrix0(z0_dm, z10_dm);
-
- if (!P10_dc) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition P_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyCell0(P0_dc, P10_dc);
- }
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t-1 = 1.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t-1 = 1.
-
-
- kalfilmsinputs_1stapp_ps->ztm1_track = 0; //Must reset to 0, meaning initial setting is done and ready for computing LH at t = 1.
-
- Update_et_Dt_1stapp(0, kalfilmsinputs_1stapp_ps);
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (1);
- }
- else
- {
- fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): calling this function makes sense only if"
- " kalfilmsinputs_1stapp_ps->ztm1_track is -1. Please check this value.");
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (0);
- }
-}
-//======================================================
-//= Integrating out the lagged regimes in order to
-//= updating zt_tm1 and Pt_tm1 for next perid tp1 through Kim-Nelson filter.
-//= tdata representing base-0 t timing, while inpt represents base-1 t timing.
-//
-//= Purpose: for each inpt, we integrate out grand regimes st
-//= only ONCE to prevent the dimension of updated zt_tm1 and Pt_tm1 through Kim-Nelson filter.
-//======================================================
-static int Updatekalfilms_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //Output:
- // tm1update
- // z_{t_1+1|t_1}
- // P_{t_1+1|t_1}
- //Input:
- // t-1: base-1 t timing. Thus t-1=inpt-1.
-
- //--- Local variables
- int stp1i, sti, t_2, t_2p1;
- double prob_previous_regimes;
- //-- Output arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Input arguments
- TSdcell *Gt_dc = kalfilmsinputs_1stapp_ps->Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //+
- TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- int T = kalfilmsinputs_1stapp_ps->T;
- TSdvector z0_sdv;
- TSdvector btp1_sdv;
- TSdvector etdata_sdv;
- //=== Work arguments.
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- //+
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //=== For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- //--- Critical checking.
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- fn_DisplayError(".../kalman.c/Updatekalfilms_1stapp(): Make sure InitializeKalman_z10_P10() is called in the function RefreshEverthing()");
-
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- btp1_sdv.n = nz;
- btp1_sdv.flag = V_DEF;
- //+
- etdata_sdv.n = ny;
- etdata_sdv.flag = V_DEF;
-
- for (t_2=kalfilmsinputs_1stapp_ps->ztm1_track; t_2<t_1; t_2++)
- {
- //If t_1 <= ztm1_track, no updating.
- //If t_1 > ztm1_track, updating z_{t|t-1} and P_{t|t-1} up to t-1 = t_1.
-
- zt_tm1_dc->C[t_2p1=t_2+1]->flag = M_GE;
- for (stp1i=nst-1; stp1i>=0; stp1i--)
- {
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //=== Updating for next period by integrating out sti..
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i], PHtran_tdata_d4->F[t_2]->C[sti], 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_d4->F[t_2]->C[sti]);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- etdata_sdv.v = etdata_dc->C[t_2]->M + ny*sti;
- z0_sdv.v = zt_tm1_dc->C[t_2]->M + nz*sti; //sti: regime at time t_2.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, &etdata_sdv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i], Pt_tm1_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
-
- //--- Integrating out the state at t_2 using
- //--- P(s_t_2|Y_{t_2}, theta) = ProbabilityStateConditionalCurrent(sti, t_2, smodel_ps);
- //--- One can also access to P(s_t_2|Y_{t_2}, theta) by using ElementV(smodel_ps->V[t_2],s_{t_2}i),
- //--- but this access will not call my function logTimetCondLH(), thus no updating for
- //--- P(s_t_2|Y_{t_2}, and thus leading to incorrect results.
- //--- we must pass t_2+1 as opposed to t_2.
- prob_previous_regimes = ProbabilityStateConditionalCurrent(sti, t_2+1, smodel_ps);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period.
- z0_sdv.v = zt_tm1_dc->C[t_2p1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[t_2p1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- //--- $$$ The following is important because it tells ProbabilityStateConditionalCurrent(), which calls
- //--- $$$ logTimetCondLH_kalfilms_1stapp(), which calls recursively this function again, that there is no
- //--- $$$ need to update Kalman filter for the period before kalfilmsinputs_1stapp_ps->ztm1_track.
- kalfilmsinputs_1stapp_ps->ztm1_track = t_2p1; //Means that z_{t_2p1+1|t_2p1} and P_{t_2p1+1|t_2p1} are done.
-
- //--- $$$ This function must be called after all the above computations are done.
- Update_et_Dt_1stapp(t_2p1, kalfilmsinputs_1stapp_ps);
- }
-
-
- //===
- DestroyMatrix_lf(Wnzbynz_dm);
- //
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (kalfilmsinputs_1stapp_ps->ztm1_track);
-}
-//======================================================
-//= Computes etdata and Dtdata for all grand regimes st at tbase0=inpt-1 or dtm1_track
-//= to prevent recomputing this object for different st at given tbase0.
-//======================================================
-static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- //Output:
- // dtm1_track is updated in this function.
- // PHtran_tdata_d4->F[t-1]
- // etdata_dc->C[t-1]
- // Dtdata_d4->F[t-1]
- //Input:
- // t_1=inpt-1: base-0 timing for et and Dt before the likelihood at time inpt is computed.
-
- //--- Local variables
- int sti, tbase0;
- //-- Output arguments
- TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdcell *yt_tm1_dc = kalfilmsinputs_1stapp_ps->yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0).
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- input arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-T.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-T.
- //+
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv;
- TSdvector yt_sdv, at_sdv;
- TSdvector etdata_sdv, yt_tm1_sdv;
- //=== Work arguments.
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- etdata_sdv.n = yt_tm1_sdv.n = ny;
- etdata_sdv.flag = yt_tm1_sdv.flag = V_DEF;
-
- for (tbase0=(kalfilmsinputs_1stapp_ps->dtm1_track+1); tbase0<=t_1; tbase0++)
- {
- //Note tbase0<=t_1, NOT tbase0<t_1.
- //If t_1 < (dtm1_track+1), no updating.
- //If t_1 >= (dtm1_track+1), updating etdata_dc->C[t-1] and Dtdata_d4->F[t-1] up to t-1=t_1.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti], 1.0, 0.0, 'N', 'T');
- CopyMatrix0(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4->F[tbase0]->C[sti], PHtran_tdata_dm);
-
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti*at_dm->nrows; //grand regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- etdata_sdv.v = etdata_dc->C[tbase0]->M + etdata_sdv.n*sti;
- yt_tm1_sdv.v = etdata_dc->C[tbase0]->M + yt_tm1_sdv.n*sti;
- CopyVector0(&yt_tm1_sdv, &at_sdv);
- MatrixTimesVector(&yt_tm1_sdv, Ht_dc->C[sti], &z0_sdv, 1.0, 1.0, 'N'); //a + H*z_{t|t-1}.
- VectorMinusVector(&etdata_sdv, &yt_sdv, &yt_tm1_sdv); //y_t - a - H*z_{t|t-1}.
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
- CopyMatrix0(Dtdata_d4->F[tbase0]->C[sti], Dtdata_dm); //Saved to be used for logTimetCondLH_kalfilms_1stapp().
- }
-
- //--- $$$ This tracker functions the same way as kalfilmsinputs_1stapp_ps->ztm1_track.
- kalfilmsinputs_1stapp_ps->dtm1_track = tbase0;
- }
-
- //===
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (kalfilmsinputs_1stapp_ps->dtm1_track);
-}
-
-
-
-
-
-
-//-----------------------------------------------------
-//------------ OLD Code --------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//
-//- IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
-//- the coefficient regime is the 1st state variable, and
-//- the volatility regime is the 2nd state variable.
-//-----------------------------------------------------
-#if defined (NEWVERSIONofDW_SWITCH)
-double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int comst_c; //composite (s_tc, s_{t-1}c)
- int st_c, stm1_c, st_v;
- int comsti_c; //composite (s_tc, s_{t-1}c)
- int sti, sti_c, stm1i_c, sti_v;
- int comstp1i_c; //composite (s_{t+1}c, s_tc)
- int stp1i, stp1i_c, stp1i_v;
- int tbase0, tp1;
- double logdet_Dtdata, loglh_timet;
- static int record_tbase1_or_inpt_or_tp1 = 0;
- static int passonce;
- double prob_previous_regimes;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRstc = kalfilmsinputs_ps->nRstc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- int **Index = smodel_ps->sv->index; //Regime-switching states.
- //smodel_ps->sv->index is for our new code.
- // For old code (before 9 April 08 and before dsge_switch is created), use smodel_ps->sv->Index;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_logTimetLH_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- tbase0 = (tp1=inpt) - 1;
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (tbase0==0)
- {
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[comsti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[comsti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, comsti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[comsti_c], Ft_dc->C[comsti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
-//=== ???????? For debugging purpose.
-//if ((inpt<2) && (st==0))
-//{
-// fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
-// fprintf(FPTR_DEBUG, "wP0_dv:\n");
-// WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Vt_dc->C[sti_v=%d]:\n", sti_v);
-// WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
-// fflush(FPTR_DEBUG);
-
-//}
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime comsti_c=%d and sti_v=%d and at time=%d\n", comsti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
- }
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- if (indxIndRegimes )
- {
- if (nRc==1) //Volatility.
- {
- comst_c = st_c = 0;
- st_v = st;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_v = Index[st][1]; //volatility state s_tv
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comst_c = st_c = st;
- st_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = st_c = Index[st][0];
- st_v = Index[st][1];
- }
- }
- else //Syncronized regimes
- {
- if (nRc>nRstc)
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = st_c;
- }
- else
- comst_c = st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[comst_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comst_c*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comst_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comst_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //--- Forming the log conditional likelihood at t.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
-//if ((inpt>82) && (inpt<86) )
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %.16e ");
-
-// fflush(FPTR_DEBUG);
-//}
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
-
-
-//=== ???????? For debugging purpose.
-if (inpt==1)
-{
- double wk1, wk2;
-
- wk1 = logdet_Dtdata;
- wk2 = VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, "logdet_Dtdata = %10.5f\n", wk1);
- fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f\n", wk2);
- fprintf(FPTR_DEBUG, "----- etdata_dv: \n");
- WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- yt_dv: \n");
- WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- at_dv: \n");
- WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- z0_dv: \n");
- WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- Ht_dc->C[comst_c=%d]:\n", comst_c);
- WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-
- fprintf(FPTR_DEBUG, "\n\n");
-
-}
-//
-fprintf(FPTR_DEBUG, " %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//fprintf(FPTR_DEBUG, "------------------------\n");
-//fprintf(FPTR_DEBUG, "st=%d and inpt=%d\n", st, inpt);
-//fprintf(FPTR_DEBUG, "loglh_timet = %10.5f\n", loglh_timet);
-//fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-//WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-////WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-////fprintf(FPTR_DEBUG, "\n");
-////WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>82) && (inpt<86) )
-//if (inpt<2)
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// tmp_Dtdata = logdeterminant(Dtdata_dm);
-// tmp_expterm = VectorDotVector(wny_dv, etdata_dv);
-// fprintf(FPTR_DEBUG, "logdeterminant(Dtdata_dm) = %10.5f;\n", tmp_Dtdata);
-// fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f;\n", tmp_expterm);
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&at_sdv:\n");
-// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&z0_sdv:\n");
-// WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Ht_dc->C[comst_c=%d]:\n",comst_c);
-// WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Rt_dc->C[st_v=%d]:\n", st_v);
-// WriteMatrix(FPTR_DEBUG, Rt_dc->C[st_v], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4->F[tbase0]->C[st = %d]:\n",st);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[st], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-
-
-
-
-//// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "zt_tm1_dc->C[tbase0]:\n");
-//// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-//// //WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-//// //fprintf(FPTR_DEBUG, "\n");
-//// fprintf(FPTR_DEBUG, "bt_dm = [\n");
-//// WriteMatrix(FPTR_DEBUG, bt_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-
-//// fprintf(FPTR_DEBUG, "et:\n");
-//// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "yt_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-//// fprintf(FPTR_DEBUG, "at_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-
-//// for (ki=0; ki<Ht_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ht_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ht_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Ft_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ft_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ft_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Vt_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Vt_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Vt_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-// fflush(FPTR_DEBUG);
-//}
-
-
- //======================================================
- //= Updating zt_tm1 and Pt_tm1 for next perid tp1.
- //= tdata = tbase0 is base-0 timing.
- //======================================================
- if (inpt > record_tbase1_or_inpt_or_tp1) //This condition always satisfies at the 1st period (which is inpt=1).
- {
- passonce = 0;
- record_tbase1_or_inpt_or_tp1 = inpt;
- }
- if (!passonce)
- {
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comstp1i_c = stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_v = Index[stp1i][1]; //volatility state s_tv
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comstp1i_c = stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comstp1i_c = stp1i_c = Index[stp1i][0];
- stp1i_v = Index[stp1i][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = stp1i_c;
- }
- else
- comstp1i_c = stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[comsti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comsti_c*at_dm->nrows; //comsti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comsti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comsti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i_v]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_{t-1}, theta) = ElementV(smodel_ps->Z[inpt],s_{inpt}_i).
- //--- Note tbase0 = inpt-1 because the data in DW code (ElementV) is base-1.
- //--- Note at this point, we cannot access to P(s_t|Y_t, theta) = ElementV(smodel_ps->V[inpt],s_{inpt}_i)
- //--- through DW's code. But we can modify my own code to do this later.
- prob_previous_regimes = ElementV(smodel_ps->Z[inpt],sti);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
- }
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>60) && (inpt<65) ) //if (inpt<5)
-//{
-// int kip1; //Must be declared at the top of this "if" block.
-
-// fprintf(FPTR_DEBUG, "zt_tm1t=[\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-
-// for (ki=0; ki<Pt_tm1_d4->F[tbase0]->ncells; ki++)
-// {
-// kip1 = ki+1;
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4t(:,:,%d)=[\n", kip1);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[ki], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-// }
-
-// fflush(FPTR_DEBUG);
-//}
-
-
-//=== ???????? For debugging purpose.
-fprintf(FPTR_DEBUG, " loglh_timet = %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-#endif
-
-
-/**
-//----------------------------------------------------------------
-//-- Tested OK, but has not use because tz_Refresh_z_T7P_T_in_kalfilms_1st_approx()
-//-- cannot access to ElementV(smodel_ps->V[tp1],sti) or ElementV(smodel_ps->V[tbase0],sti)
-//-- because no likelihood has been formed at all before this function is called.
-//----------------------------------------------------------------
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//-----------------------------------------------------
-void tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- double debug1;
- //--- Local variables
- int stp1i, stp1i_c, stp1i_v, sti, sti_c, sti_v, tbase0, tp1;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, sti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti_c], Ft_dc->C[sti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime sti_c=%d and sti_v=%d and at time=%d\n", sti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
-
-
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[0]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[0]->C[0], " %.16e ");
-
-
- //============== Updating zt_tm1 and Pt_tm1. ==================
- for (tbase0=0; tbase0<T; tbase0++ )
- {
- //tdata = tbase0 is base-0 timing.
- tp1 = tbase0 + 1; //Next period.
-
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if (indxIndRegimes)
- {
- stp1i_c = smodel_ps->sv->Index[stp1i][0];
- stp1i_v = smodel_ps->sv->Index[stp1i][1];
- }
- else
- {
- stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti_c*at_dm->nrows; //sti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[sti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_t, theta) = ElementV(smodel_ps->V[t+1],s_{t+1}_i).
- //--- Note because the data in DW code (ElementV) is base-1, t+1 is actually tbase0.
- debug1 = ElementV(smodel_ps->V[tp1],sti); //?????? Debug.
- //ScalarTimesVectorUpdate(ztp1_dv, ElementV(smodel_ps->V[tp1],sti), ztp1_t_dv);
- //ScalarTimesMatrix(Ptp1_dm, ElementV(smodel_ps->V[tp1],sti), Ptp1_t_dm, 1.0);
- ScalarTimesVectorUpdate(ztp1_dv, 0.5, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, 0.5, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
-
-// fprintf(FPTR_DEBUG, "\n &yt_sdv:\n");
-// WriteMatrix(FPTR_DEBUG, &yt_sdv, " %.16e ");
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[tp1]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tp1], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[tp1]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tp1]->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fflush(FPTR_DEBUG);
-
-
- }
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-}
-//-----------------------------------------------------
-//- Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// (1) RunningGensys_const7varionly(lwzmodel_ps);
-// (2) Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-// (3) tz_Refresh_z_T7P_T_in_kalfilms_1st_approx();
-//- before using kalfilms_timet_1st_approx().
-//-----------------------------------------------------
-double tz_kalfilms_timet_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0. The same for (T+1)-by-1 gbeta_dv and nlcoefs-by-(T+1) galpha_dm.
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int st_c, st_v, tbase0;
- double loglh_timet;
- //--- Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- TSdvector yt_sdv, at_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- if (smodel_ps->sv->nstates != zt_tm1_dc->C[0]->ncols) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Make sure that the column dimension of zt_tm1_dc->C is the same as smodel_ps->sv->nstates");
-
- tbase0 = inpt - 1; //base-0 time t.
-
- if (indxIndRegimes && (nRc==1))
- {
- st_c = 0;
- st_v = st;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- st_c = st;
- st_v = 0;
- }
- else if (indxIndRegimes)
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
- st_c = smodel_ps->sv->Index[st][0];
- st_v = smodel_ps->sv->Index[st][1];
- }
- else
- {
- st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + st_c*at_dm->nrows; //st_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[st_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
-
- //--- Forming the log conditional likelihood at t.
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //===
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-/**/
-
-
-
-
+/*===============================================================================================================
+ * Check $$$ for important notes.
+ * Check <<>> for updating DW's new switch code or questions for DW.
+ *
+ * kalcvf_urw(): the Kalman filter forward prediction specialized for only a univariate random walk (urw) process.
+ *
+ * State space model is defined as follows:
+ * z(t+1) = z(t)+eta(t) (state or transition equation)
+ * y(t) = x(t)'*z(t)+eps(t) (observation or measurement equation)
+ * where for this function, eta and eps must be uncorrelated; y(t) must be 1-by-1. Note that
+ * x(t): k-by-1;
+ * z(t): k-by-1;
+ * eps(t): 1-by-1 and ~ N(0, sigma^2);
+ * eta(t): ~ N(0, V) where V is a k-by-k covariance matrix.
+ *
+ *
+ * Written by Tao Zha, May 2004.
+ * Revised, May 2008;
+=================================================================================================================*/
+
+/**
+//=== For debugging purpose.
+if (1)
+{
+ double t_loglht;
+
+ t_loglht = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
+ fprintf(FPTR_DEBUG, " %10.5f\n", t_loglht);
+
+ fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
+
+ fprintf(FPTR_DEBUG, "\n wP0_dv:\n");
+ WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
+ fprintf(FPTR_DEBUG, "\n Vt_dc->C[sti_v=%d]:\n", sti_v);
+ WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
+
+ fflush(FPTR_DEBUG);
+}
+/**/
+
+
+#include "kalman.h"
+
+
+static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
+int Updatekalfilms_1stapp(int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
+
+
+TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv) //, int storeZ, int storeV)
+{
+ int _i;
+ //===
+ TSivector *rows_iv = NULL;
+ TSivector *cols_iv = NULL;
+ //---
+ TSkalcvfurw *kalcvfurw_ps = tzMalloc(1, TSkalcvfurw);
+
+
+ kalcvfurw_ps->indx_tvsigmasq = tv;
+ kalcvfurw_ps->fss = T;
+ kalcvfurw_ps->kx = k;
+
+ //===
+ kalcvfurw_ps->V_dm = CreateMatrix_lf(k, k);
+ kalcvfurw_ps->ylhtran_dv = CreateVector_lf(T);
+ kalcvfurw_ps->Xrhtran_dm = CreateMatrix_lf(k, T);
+ kalcvfurw_ps->z10_dv = CreateVector_lf(k);
+ kalcvfurw_ps->P10_dm = CreateMatrix_lf(k, k);
+
+ kalcvfurw_ps->zupdate_dv = CreateVector_lf(k);
+ kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
+ kalcvfurw_ps->ylhtranpred_dv = CreateVector_lf(T);
+ //
+ rows_iv = CreateVector_int(T);
+ cols_iv = CreateVector_int(T);
+ for (_i=T-1; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
+ kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
+ // if (!storeZ) kalcvfurw_ps->Zpredtran_dm = (TSdmatrix *)NULL;
+ // else kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
+ // if (!storeV) kalcvfurw_ps->Ppred_dc = (TSdcell *)NULL;
+ // else {
+ // rows_iv = CreateVector_int(T);
+ // cols_iv = CreateVector_int(T);
+ // for (_i=T; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
+ // kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
+ // }
+
+ DestroyVector_int(rows_iv);
+ DestroyVector_int(cols_iv);
+ return (kalcvfurw_ps);
+}
+
+TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps)
+{
+ if (kalcvfurw_ps) {
+ DestroyMatrix_lf(kalcvfurw_ps->V_dm);
+ DestroyVector_lf(kalcvfurw_ps->ylhtran_dv);
+ DestroyMatrix_lf(kalcvfurw_ps->Xrhtran_dm);
+ DestroyVector_lf(kalcvfurw_ps->z10_dv);
+ DestroyMatrix_lf(kalcvfurw_ps->P10_dm);
+
+ DestroyVector_lf(kalcvfurw_ps->zupdate_dv);
+ DestroyMatrix_lf(kalcvfurw_ps->Zpredtran_dm);
+ DestroyCell_lf(kalcvfurw_ps->Ppred_dc);
+ DestroyVector_lf(kalcvfurw_ps->ylhtranpred_dv);
+
+ free(kalcvfurw_ps);
+ return ((TSkalcvfurw *)NULL);
+ }
+ else return (kalcvfurw_ps);
+}
+
+
+void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps)
+{
+ //See the notes of SWZ regarding the government's updating of the parameters in their Phillips-curve equation.
+ //NOTE: make sure that the value of kalcvfurw_ps->sigmasq and other input values are given.
+ int ti;
+ double workd, workdenominv;
+ //---
+ int fss, kx;
+ double sigmasq_fix = kalcvfurw_ps->sigmasq;
+// double sigmasq;
+ TSdmatrix *V_dm;
+ TSdmatrix *Zpredtran_dm;
+ TSdcell *Ppred_dc;
+ TSdvector *ylhtran_dv;
+ TSdmatrix *Xrhtran_dm;
+ //===
+ TSdvector *workkxby1_dv = NULL; //kx-by-1.
+// TSdvector *work1kxby1_dv = NULL; //kx-by-1.
+ TSdmatrix *workkxbykx_dm = NULL; //kx-by-kx symmetric and positive positive.
+// //===
+// TSdvector *zbefore_dv = CreateVector_lf(kalcvfurw_ps->kx);
+// TSdmatrix *Vbefore_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
+// TSdvector *zafter_dv = CreateVector_lf(kalcvfurw_ps->kx);
+// TSdmatrix *Vafter_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
+ //******* WARNING: Some dangerous pointer movement to gain efficiency *******
+// double *yt_p;
+// double *Vbefore_p;
+// double *Vafter_p;
+ TSdvector xt_sdv;
+ TSdvector zbefore_sdv;
+ //TSdmatrix Vbefore_sdm;
+ TSdvector zafter_sdv;
+ //TSdmatrix Vafter_sdm;
+
+
+ if (!kalcvfurw_ps) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps must be created");
+ if (!kalcvfurw_ps->V_dm || !kalcvfurw_ps->ylhtran_dv || !kalcvfurw_ps->Xrhtran_dm || !kalcvfurw_ps->z10_dv || !kalcvfurw_ps->P10_dm)
+ fn_DisplayError(".../kalcvf_urw(): input arguments kalcvfurw_ps->V_dm, kalcvfurw_ps->ylhtran_dv, kalcvfurw_ps->Xrhtran_dm, kalcvfurw_ps->z10_dv, kalcvfurw_ps->P10_dm must be given legal values");
+ if (!(kalcvfurw_ps->P10_dm->flag & (M_SU | M_SL))) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps->P10_dm must be symmetric");
+ fss = kalcvfurw_ps->fss;
+ kx = kalcvfurw_ps->kx;
+ V_dm = kalcvfurw_ps->V_dm;
+ Zpredtran_dm = kalcvfurw_ps->Zpredtran_dm;
+ Ppred_dc = kalcvfurw_ps->Ppred_dc;
+ ylhtran_dv = kalcvfurw_ps->ylhtran_dv;
+ Xrhtran_dm = kalcvfurw_ps->Xrhtran_dm;
+ //---
+ xt_sdv.n = kx;
+ xt_sdv.flag = V_DEF;
+ zbefore_sdv.n = kx;
+ zbefore_sdv.flag = V_DEF;
+ zafter_sdv.n = kx;
+ zafter_sdv.flag = V_DEF;
+
+ //=== Memory allocation.
+ workkxby1_dv = CreateVector_lf(kx);
+ workkxbykx_dm = CreateMatrix_lf(kx, kx);
+
+
+ //------- The first period (ti=0). -------
+ zbefore_sdv.v = kalcvfurw_ps->z10_dv->v;
+ zafter_sdv.v = Zpredtran_dm->M;
+ xt_sdv.v = Xrhtran_dm->M;
+ //---
+
+ workd = ylhtran_dv->v[0] - (kalcvfurw_ps->ylhtranpred_dv->v[0]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
+ SymmatrixTimesVector(workkxby1_dv, kalcvfurw_ps->P10_dm, &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
+
+ if (!kalcvfurw_ps->indx_tvsigmasq)
+ workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
+ else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
+ workdenominv = 1.0/(sigmasq_fix*square(kalcvfurw_ps->z10_dv->v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t];
+ else {
+ printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
+ exit(EXIT_FAILURE);
+ }
+
+
+ //--- Updating z_{t+1|t}.
+ CopyVector0(&zafter_sdv, &zbefore_sdv);
+ VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
+ //--- Updating P_{t+1|t}.
+ CopyMatrix0(workkxbykx_dm, V_dm);
+ VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
+ // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
+ MatrixPlusMatrix(Ppred_dc->C[0], kalcvfurw_ps->P10_dm, workkxbykx_dm);
+ //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
+ Ppred_dc->C[0]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
+ // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
+ //Done with all work* arrays.
+
+ //------- The rest of the periods (ti=1:T-1). -------
+ for (ti=1; ti<fss; ti++) {
+ //NOTE: ti=0 has been taken care of outside of this loop.
+ zbefore_sdv.v = Zpredtran_dm->M + (ti-1)*kx;
+ zafter_sdv.v = Zpredtran_dm->M + ti*kx;
+ xt_sdv.v = Xrhtran_dm->M + ti*kx;
+ //---
+ workd = ylhtran_dv->v[ti] - (kalcvfurw_ps->ylhtranpred_dv->v[ti]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
+ SymmatrixTimesVector(workkxby1_dv, Ppred_dc->C[ti-1], &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
+ if (!kalcvfurw_ps->indx_tvsigmasq)
+ workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
+ else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
+ workdenominv = 1.0/(sigmasq_fix*square(zbefore_sdv.v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
+ else {
+ printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
+ exit(EXIT_FAILURE);
+ }
+ //--- Updating z_{t+1|t}.
+ CopyVector0(&zafter_sdv, &zbefore_sdv);
+ VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
+ //--- Updating P_{t+1|t}.
+ CopyMatrix0(workkxbykx_dm, V_dm);
+ VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
+ // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
+ MatrixPlusMatrix(Ppred_dc->C[ti], Ppred_dc->C[ti-1], workkxbykx_dm);
+ //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
+ Ppred_dc->C[ti]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
+ // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
+ //Done with all work* arrays.
+ }
+ CopyVector0(kalcvfurw_ps->zupdate_dv, &zafter_sdv);
+ Zpredtran_dm->flag = M_GE;
+ kalcvfurw_ps->ylhtranpred_dv->flag = V_DEF;
+
+// DestroyVector_lf(zbefore_dv);
+// DestroyMatrix_lf(Vbefore_dm);
+// DestroyVector_lf(zafter_dv);
+// DestroyMatrix_lf(Vafter_dm);
+
+ DestroyVector_lf(workkxby1_dv);
+// DestroyVector_lf(work1kxby1_dv);
+ DestroyMatrix_lf(workkxbykx_dm);
+}
+
+
+
+//-----------------------------------------------------------------------------------------------------------------------
+//-- General constant (known-time-varying) Kalman filter for DSGE models.
+//-----------------------------------------------------------------------------------------------------------------------
+struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T)
+{
+ int _i;
+ //===
+ TSivector *rows_iv = CreateVector_int(T);
+ TSivector *cols_iv = CreateVector_int(T);
+ //~~~ Creating the structure and initializing the NULL pointers.
+ struct TSkalfiltv_tag *kalfiltv_ps = tzMalloc(1, struct TSkalfiltv_tag);
+
+
+ //--- Default value.
+ kalfiltv_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ //0: using the unconditional mean for any given regime at time 0.
+ //--- Other assignments.
+ kalfiltv_ps->ny = ny;
+ kalfiltv_ps->nz = nz;
+ kalfiltv_ps->T = T;
+
+
+
+ //--------- Creates memory and assigns values. The order matters.
+ kalfiltv_ps->yt_dm = CreateMatrix_lf(ny, T);
+ kalfiltv_ps->at_dm = CreateMatrix_lf(ny, T);
+ //
+ for (_i=T-1; _i>=0; _i--)
+ {
+ rows_iv->v[_i] = ny;
+ cols_iv->v[_i] = nz;
+ }
+ rows_iv->flag = cols_iv->flag = V_DEF;
+ kalfiltv_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
+ //
+ for (_i=T-1; _i>=0; _i--)
+ {
+ rows_iv->v[_i] = ny;
+ cols_iv->v[_i] = ny;
+ }
+ kalfiltv_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
+ //
+ for (_i=T-1; _i>=0; _i--)
+ {
+ rows_iv->v[_i] = nz;
+ cols_iv->v[_i] = ny;
+ }
+ kalfiltv_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
+ //
+ kalfiltv_ps->bt_dm = CreateMatrix_lf(nz, T);
+ //
+ for (_i=T-1; _i>=0; _i--)
+ {
+ rows_iv->v[_i] = nz;
+ cols_iv->v[_i] = nz;
+ }
+ kalfiltv_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
+ kalfiltv_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
+ //
+ kalfiltv_ps->z0_dv = CreateVector_lf(nz);
+ kalfiltv_ps->P0_dm = CreateMatrix_lf(nz, nz);
+
+
+ //---
+ kalfiltv_ps->zt_tm1_dm = CreateMatrix_lf(nz, T);
+ for (_i=T-1; _i>=0; _i--)
+ {
+ rows_iv->v[_i] = nz;
+ cols_iv->v[_i] = nz;
+ }
+ kalfiltv_ps->Pt_tm1_dc = CreateCell_lf(rows_iv, cols_iv);
+
+
+ //===
+ DestroyVector_int(rows_iv);
+ DestroyVector_int(cols_iv);
+
+ return (kalfiltv_ps);
+
+}
+//---
+struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
+{
+ if (kalfiltv_ps)
+ {
+ //=== The order matters!
+ DestroyMatrix_lf(kalfiltv_ps->yt_dm);
+ DestroyMatrix_lf(kalfiltv_ps->at_dm);
+ DestroyCell_lf(kalfiltv_ps->Ht_dc);
+ DestroyCell_lf(kalfiltv_ps->Rt_dc);
+ DestroyCell_lf(kalfiltv_ps->Gt_dc);
+ //---
+ DestroyMatrix_lf(kalfiltv_ps->bt_dm);
+ DestroyCell_lf(kalfiltv_ps->Ft_dc);
+ DestroyCell_lf(kalfiltv_ps->Vt_dc);
+ //---
+ DestroyVector_lf(kalfiltv_ps->z0_dv);
+ DestroyMatrix_lf(kalfiltv_ps->P0_dm);
+ //---
+ DestroyMatrix_lf(kalfiltv_ps->zt_tm1_dm);
+ DestroyCell_lf(kalfiltv_ps->Pt_tm1_dc);
+
+
+ //---
+ tzDestroy(kalfiltv_ps); //Must be freed last!
+
+ return ((struct TSkalfiltv_tag *)NULL);
+ }
+ else return (kalfiltv_ps);
+};
+
+
+//-----------------------------------------------------------------------------------------------------------------------
+//-- New code: Inputs for filter for Markov-switching DSGE models at any time t.
+//-----------------------------------------------------------------------------------------------------------------------
+struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp2(int ny, int nz, int nu, int ne, int nst, int T)
+{
+ //~~~ Creating the structure and initializing the NULL pointers.
+ struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
+
+ //===
+ TSivector *rows_iv = NULL;
+ TSivector *cols_iv = NULL;
+
+ //=== Default value.
+ kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ //0: using the unconditional mean for any given regime at time 0.
+ kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
+ //0: using the unconditional moments.
+ kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
+ kalfilmsinputs_1stapp_ps->ztm1_track = -1;
+ kalfilmsinputs_1stapp_ps->dtm1_track = -1;
+
+ //--- Other key assignments.
+ kalfilmsinputs_1stapp_ps->ny = ny;
+ kalfilmsinputs_1stapp_ps->nz = nz;
+ kalfilmsinputs_1stapp_ps->nu = nu;
+ kalfilmsinputs_1stapp_ps->ne = ne;
+ kalfilmsinputs_1stapp_ps->nst = nst;
+ kalfilmsinputs_1stapp_ps->T = T;
+
+ //--------- Creates memory and assigns values. The order matters.
+ kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
+ kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
+ //
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ if (nu)
+ {
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, nu);
+ kalfilmsinputs_1stapp_ps->Psiut_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ }
+ else
+ kalfilmsinputs_1stapp_ps->Psiut_dc = NULL;
+ //
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, ne);
+ kalfilmsinputs_1stapp_ps->Psiet_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
+ kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+
+ //--- For output arguments.
+ rows_iv = CreateConstantVector_int(T+1, nz);
+ cols_iv = CreateConstantVector_int(T+1, nst);
+ kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(T, ny);
+ cols_iv = CreateConstantVector_int(T, nst);
+ kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
+ rows_iv = CreateConstantVector_int(T, ny);
+ cols_iv = CreateConstantVector_int(T, nst);
+ //
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+
+ return (kalfilmsinputs_1stapp_ps);
+}
+//---
+struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp2(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
+{
+ if (kalfilmsinputs_1stapp_ps)
+ {
+ //=== The order matters!
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Psiut_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
+ //---
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Psiet_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
+ //---
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
+ //---
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
+ DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
+ DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
+ DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
+ //---
+ tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
+
+ return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
+ }
+ else return (kalfilmsinputs_1stapp_ps);
+};
+
+struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T)
+{
+ //~~~ Creating the structure and initializing the NULL pointers.
+ struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
+
+ //===
+ TSivector *rows_iv = NULL;
+ TSivector *cols_iv = NULL;
+
+ //=== Default value.
+ kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ //0: using the unconditional mean for any given regime at time 0.
+ kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
+ //0: using the unconditional moments.
+ kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
+ kalfilmsinputs_1stapp_ps->ztm1_track = -1;
+ kalfilmsinputs_1stapp_ps->dtm1_track = -1;
+
+ //--- Other key assignments.
+ kalfilmsinputs_1stapp_ps->ny = ny;
+ kalfilmsinputs_1stapp_ps->nz = nz;
+ kalfilmsinputs_1stapp_ps->nst = nst;
+ kalfilmsinputs_1stapp_ps->T = T;
+
+ //--------- Creates memory and assigns values. The order matters.
+ kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
+ kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
+ //
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
+ kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+
+ //--- For output arguments.
+ rows_iv = CreateConstantVector_int(T+1, nz);
+ cols_iv = CreateConstantVector_int(T+1, nst);
+ kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, nz);
+ kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nst, nz);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(T, ny);
+ cols_iv = CreateConstantVector_int(T, nst);
+ kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
+ rows_iv = CreateConstantVector_int(T, ny);
+ cols_iv = CreateConstantVector_int(T, nst);
+ //
+ rows_iv = CreateConstantVector_int(nst, ny);
+ cols_iv = CreateConstantVector_int(nst, ny);
+ kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+
+ return (kalfilmsinputs_1stapp_ps);
+}
+//---
+struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
+{
+ if (kalfilmsinputs_1stapp_ps)
+ {
+ //=== The order matters!
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
+ //---
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
+ //---
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
+ DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
+ //---
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
+ DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
+ DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
+ DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
+ DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
+ //---
+ tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
+
+ return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
+ }
+ else return (kalfilmsinputs_1stapp_ps);
+};
+
+
+//-----------------------------------------------------------------------------------------------------------------------
+//-- OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
+//-----------------------------------------------------------------------------------------------------------------------
+struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T)
+{
+ //~~~ Creating the structure and initializing the NULL pointers.
+ struct TSkalfilmsinputs_tag *kalfilmsinputs_ps = tzMalloc(1, struct TSkalfilmsinputs_tag);
+
+ //===
+ TSivector *rows_iv = NULL;
+ TSivector *cols_iv = NULL;
+
+ //--- Default value.
+ kalfilmsinputs_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ //0: using the unconditional mean for any given regime at time 0.
+ //--- Other assignments.
+ kalfilmsinputs_ps->ny = ny;
+ kalfilmsinputs_ps->nz = nz;
+ kalfilmsinputs_ps->nRc = nRc;
+ kalfilmsinputs_ps->nRstc = nRstc;
+ kalfilmsinputs_ps->nRv = nRv;
+ kalfilmsinputs_ps->indxIndRegimes = indxIndRegimes;
+ kalfilmsinputs_ps->T = T;
+
+
+ //--------- Creates memory and assigns values. The order matters.
+ kalfilmsinputs_ps->yt_dm = CreateMatrix_lf(ny, T);
+ kalfilmsinputs_ps->at_dm = CreateMatrix_lf(ny, nRc);
+ //
+ rows_iv = CreateConstantVector_int(nRc, ny);
+ cols_iv = CreateConstantVector_int(nRc, nz);
+ kalfilmsinputs_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nRv, ny);
+ cols_iv = CreateConstantVector_int(nRv, ny);
+ kalfilmsinputs_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nRv, nz);
+ cols_iv = CreateConstantVector_int(nRv, ny);
+ kalfilmsinputs_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ kalfilmsinputs_ps->bt_dm = CreateMatrix_lf(nz, nRc);
+ //
+ rows_iv = CreateConstantVector_int(nRc, nz);
+ cols_iv = CreateConstantVector_int(nRc, nz);
+ kalfilmsinputs_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nRv, nz);
+ cols_iv = CreateConstantVector_int(nRv, nz);
+ kalfilmsinputs_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ if (indxIndRegimes)
+ {
+ kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRc*nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
+ //
+ rows_iv = CreateConstantVector_int(nRc*nRv, nz);
+ cols_iv = CreateConstantVector_int(nRc*nRv, nz);
+ kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ }
+ else
+ {
+ if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
+ kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
+ //
+ rows_iv = CreateConstantVector_int(nRv, nz);
+ cols_iv = CreateConstantVector_int(nRv, nz);
+ kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ }
+ //--- For output arguments.
+ if (indxIndRegimes)
+ {
+ rows_iv = CreateConstantVector_int(T, nz);
+ cols_iv = CreateConstantVector_int(T, nRc*nRv);
+ kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nRc*nRv, nz);
+ cols_iv = CreateConstantVector_int(nRc*nRv, nz);
+ kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ }
+ else
+ {
+ if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
+ rows_iv = CreateConstantVector_int(T, nz);
+ cols_iv = CreateConstantVector_int(T, nRv);
+ kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ //
+ rows_iv = CreateConstantVector_int(nRv, nz);
+ cols_iv = CreateConstantVector_int(nRv, nz);
+ kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ rows_iv = DestroyVector_int(rows_iv);
+ cols_iv = DestroyVector_int(cols_iv);
+ }
+
+
+ //===
+ DestroyVector_int(rows_iv);
+ DestroyVector_int(cols_iv);
+
+ return (kalfilmsinputs_ps);
+
+}
+//---
+struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps)
+{
+ if (kalfilmsinputs_ps)
+ {
+ //=== The order matters!
+ DestroyMatrix_lf(kalfilmsinputs_ps->yt_dm);
+ DestroyMatrix_lf(kalfilmsinputs_ps->at_dm);
+ DestroyCell_lf(kalfilmsinputs_ps->Ht_dc);
+ DestroyCell_lf(kalfilmsinputs_ps->Rt_dc);
+ DestroyCell_lf(kalfilmsinputs_ps->Gt_dc);
+ //---
+ DestroyMatrix_lf(kalfilmsinputs_ps->bt_dm);
+ DestroyCell_lf(kalfilmsinputs_ps->Ft_dc);
+ DestroyCell_lf(kalfilmsinputs_ps->Vt_dc);
+ //---
+ DestroyMatrix_lf(kalfilmsinputs_ps->z0_dm);
+ DestroyCell_lf(kalfilmsinputs_ps->P0_dc);
+ //---
+ DestroyCell_lf(kalfilmsinputs_ps->zt_tm1_dc);
+ DestroyFourth_lf(kalfilmsinputs_ps->Pt_tm1_d4);
+ //---
+ tzDestroy(kalfilmsinputs_ps); //Must be freed last!
+
+ return ((struct TSkalfilmsinputs_tag *)NULL);
+ }
+ else return (kalfilmsinputs_ps);
+};
+
+
+#define LOG2PI (1.837877066409345e+000) //log(2*pi)
+//-----------------------------------------------------
+//-- Constant-parameters (known-time-varying) Kalman filter
+//-----------------------------------------------------
+double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
+{
+ //General constant (known-time-varying) Kalman filter for DSGE models (conditional on all the parameters).
+ // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
+ // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
+ //
+ // State space model is defined as follows:
+ // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
+ // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
+ // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
+ //
+ // Inputs are as follows:
+ // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
+ // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
+ // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
+ // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
+ // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
+ // ------
+ // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
+ // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
+ // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
+ // ------
+ // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ // 0: using the unconditional mean for any given regime at time 0.
+ // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Do not enter if indxIni=0.)
+ // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Do not enter if indxIni=0.)
+ //
+ // Outputs are as follows:
+ // loglh is a value of the log likelihood function of the state-space model
+ // under the assumption that errors are multivariate Gaussian.
+ // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
+ // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
+ // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
+ // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
+ //
+ // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
+ // z0_0m1 = (I-F(:,:,1))\b(:,1)
+ // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
+ // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
+ //
+ // March 2007, written by Tao Zha
+ // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
+
+ int T = kalfiltv_ps->T;
+ int Tp1 = T + 1;
+ int ny = kalfiltv_ps->ny;
+ int nz = kalfiltv_ps->nz;
+ int indx_badlh = 0; //1: bad likelihood with, say, -infinity of the LH value.
+ int tdata, ti;
+ //--- Work arguments.
+ int nz2 = square(nz);
+ TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdvector *wP0_dv = CreateVector_lf(nz2);
+ //+
+ TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv; //double loglh_tdata; //logdetDtdata.
+ TSdvector *wny_dv = CreateVector_lf(ny);
+ TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
+ TSdvector *etdata_dv = CreateVector_lf(ny);
+ TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
+ TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
+ //--- For eigenvalue decompositions
+ int ki;
+ int errflag;
+ double eigmax, logdet_Dtdata;
+ TSdzvector *evals_dzv = NULL;
+ TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
+ //--- Input arguments.
+ TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
+ TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
+ TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
+ //
+ TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
+ TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
+ TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
+ //
+ TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
+ TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
+ //--- Output arguments.
+ double loglh; //log likelihood.
+ TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
+ TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
+
+
+
+ //=== Initializing.
+ if (!kalfiltv_ps->indxIni)
+ {
+ InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
+ InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
+
+ //=== Eigenanalysis to determine the roots to ensure boundedness.
+ evals_dzv = CreateVector_dz(nz);
+ evals_abs_dv = CreateVector_lf(nz);
+ errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
+ if (errflag) fn_DisplayError("tz_kalfiltv() in kalman.c: eigen decomposition failed");
+ for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
+ evals_abs_dv->flag = V_DEF;
+ eigmax = MaxVector(evals_abs_dv);
+ if (eigmax < (1.0+1.0e-14))
+ {
+ //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
+ MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
+ CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
+ bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
+ //Done with Wnzbynz_dm.
+ //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
+ tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
+ MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
+ CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
+ bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
+ CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
+ //Done with all w*_dv and W*_dm.
+ }
+ else
+ {
+ fprintf(stdout, "Fatal error: tz_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
+ " and the initial conditions must be supplied by, say, input arguments");
+ fflush(stdout);
+ exit( EXIT_FAILURE );
+ }
+ }
+ CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
+ CopyMatrix0(Pt_tm1_dc->C[0], P0_dm);
+
+ //====== See p.002 in LiuWZ. ======
+ at_sdv.n = yt_sdv.n = yt_dm->nrows;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+ zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
+ zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
+ btp1_sdv.n = bt_dm->nrows;
+ btp1_sdv.flag = V_DEF;
+ loglh = 0.0;
+ for (tdata=0; tdata<T; tdata++ )
+ {
+ //Base-0 timing.
+ ti = tdata + 1; //Next period.
+
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
+ yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
+ at_sdv.v = at_dm->M + tdata*at_dm->nrows;
+ zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
+ VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
+ MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
+ //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
+ // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
+ // a bad number or a complex number.
+ Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
+
+ //--- Forming the log likelihood.
+ if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (kalfiltv_ps->loglh = -NEARINFINITY);
+ bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
+ loglh += -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
+ //loglh += -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
+ //Done with all w*_dv.
+
+
+ //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
+ if (ti<T)
+ {
+ //Updating only up to tdata=T-2. The values at ti=T or tdata=T-1 will not be used in the likelihood function.
+
+ //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
+ CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
+ MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[ti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
+ //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
+ ztp1_t_sdv.v = zt_tm1_dm->M + ti*zt_tm1_dm->nrows;
+ MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[ti], &zt_tm1_sdv, 1.0, 0.0, 'N');
+ MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
+ btp1_sdv.v = bt_dm->M + ti*btp1_sdv.n;
+ VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
+ //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
+ CopyMatrix0(Pt_tm1_dc->C[ti], Vt_dc->C[ti]);
+ MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
+ MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[ti], Wnzbynz_dm, -1.0);
+ //Done with all W*_dm.
+ MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[ti], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[ti], 1.0, 0.0, 'N', 'T');
+ MatrixPlusMatrixUpdate(Pt_tm1_dc->C[ti], W2nzbynz_dm);
+ //Done with all W*_dm.
+ }
+ }
+ zt_tm1_dm->flag = M_GE;
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+ //
+ DestroyVector_lf(wny_dv);
+ DestroyMatrix_lf(Wnzbyny_dm);
+ DestroyMatrix_lf(W2nzbynz_dm);
+ DestroyMatrix_lf(PHtran_tdata_dm);
+ DestroyVector_lf(etdata_dv);
+ DestroyMatrix_lf(Dtdata_dm);
+ DestroyMatrix_lf(Kt_tdata0_dm);
+ DestroyMatrix_lf(Kt_tdata_dm);
+
+ return (kalfiltv_ps->loglh = loglh);
+}
+/**
+double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
+{
+ //This function is used to test tz_logTimetCondLH_kalfiltv().
+ int T = kalfiltv_ps->T;
+ int tdata;
+ double loglh;
+
+ loglh = 0.0;
+ for (tdata=0; tdata<T; tdata++) loglh += tz_logTimetCondLH_kalfiltv(0, tdata+1, kalfiltv_ps);
+
+ return (loglh);
+}
+/**/
+//-----------------------------------------------------
+//-- Updating Kalman filter at time t for constant-parameters (or known-time-varying) Kalman filter.
+//-----------------------------------------------------
+double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps)
+{
+ //st: base-0 grand regime at time t, which is just a dummy for this constant-parameter function in order to use
+ // Waggoner's automatic functions.
+ //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
+ // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
+ // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
+ //See (42.3) on p.42 in the SWZII NOTES.
+ //
+ //log LH at time t for constant (known-time-varying) Kalman-filter DSGE models (conditional on all the parameters).
+ // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood at time t.
+ // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
+ //
+ // State space model is defined as follows:
+ // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
+ // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
+ // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
+ //
+ // Inputs are as follows:
+ // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
+ // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
+ // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
+ // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
+ // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
+ // ------
+ // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
+ // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
+ // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
+ // ------
+ // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ // 0: using the unconditional mean for any given regime at time 0.
+ // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
+ // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
+ //
+ // Outputs are as follows:
+ // loglh is a value of the log likelihood function of the state-space model
+ // under the assumption that errors are multivariate Gaussian.
+ // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
+ // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
+ // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
+ // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
+ //
+ // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
+ // z0_0m1 = (I-F(:,:,1))\b(:,1)
+ // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
+ // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
+ //
+ // April 2008, written by Tao Zha
+ // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
+
+ //--- Output arguments.
+ double loglh_timet; //log likelihood at time t.
+ TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
+ TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
+ //--- Input arguments.
+ int tdata, tp1;
+ TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
+ TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
+ int T = kalfiltv_ps->T;
+ int ny = kalfiltv_ps->ny;
+ int nz = kalfiltv_ps->nz;
+ //--- Work arguments.
+ int nz2 = square(nz);
+ TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdvector *wP0_dv = CreateVector_lf(nz2);
+ //+
+ TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv;
+ TSdvector *wny_dv = CreateVector_lf(ny);
+ TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
+ TSdvector *etdata_dv = CreateVector_lf(ny);
+ TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
+ TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
+ //--- For eigenvalue decompositions
+ int ki;
+ int errflag;
+ double eigmax, logdet_Dtdata;
+ TSdzvector *evals_dzv = NULL;
+ TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
+ //--- Input arguments.
+ TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
+ TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
+ TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
+ //
+ TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
+ TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
+ TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
+ //
+
+
+ tdata = (tp1=inpt) - 1; //Base-0 time.
+
+ //======= Initial condition. =======
+ if (tdata==0)
+ {
+ //=== Initializing.
+ if (!kalfiltv_ps->indxIni)
+ {
+ InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
+ InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
+
+ //=== Eigenanalysis to determine the roots to ensure boundedness.
+ evals_dzv = CreateVector_dz(nz);
+ evals_abs_dv = CreateVector_lf(nz);
+ errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
+ if (errflag) fn_DisplayError("tz_logTimetCondLH_kalfiltv() in kalman.c: eigen decomposition failed");
+ for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
+ evals_abs_dv->flag = V_DEF;
+ eigmax = MaxVector(evals_abs_dv);
+ if (eigmax < (1.0+1.0e-14))
+ {
+ //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
+ MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
+ CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
+ bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
+ //Done with Wnzbynz_dm.
+ //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
+ tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
+ MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
+ CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
+ bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
+ CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
+ //Done with all w*_dv and W*_dm.
+ }
+ else
+ {
+ fprintf(FPTR_DEBUG, "Fatal error: tz_logTimetCondLH_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
+ " and thus the initial conditions must be supplied by, say, input arguments");
+ fflush(FPTR_DEBUG);
+ exit( EXIT_FAILURE );
+ }
+ }
+ CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
+ CopyMatrix0(Pt_tm1_dc->C[tdata], P0_dm);
+ }
+
+
+ //======= Liklihood at time t (see p.002 in LiuWZ). =======
+ at_sdv.n = yt_sdv.n = yt_dm->nrows;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+ zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
+ zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
+ btp1_sdv.n = bt_dm->nrows;
+ btp1_sdv.flag = V_DEF;
+
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
+ yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
+ at_sdv.v = at_dm->M + tdata*at_dm->nrows;
+ zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
+ VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
+ MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
+ //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
+ // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
+ // a bad number or a complex number.
+ Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
+
+ //--- Forming the log likelihood.
+ if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
+ bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
+ loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
+ //Done with all w*_dv.
+
+
+ //======= Updating for the next period. =======
+ //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
+ if (tp1<T)
+ {
+ //Updating only up to tdata=T-2, because the values at tp1=T or tdata=T-1 will NOT be used in the likelihood function.
+
+ //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
+ CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
+ MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[tp1], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
+ //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
+ ztp1_t_sdv.v = zt_tm1_dm->M + tp1*zt_tm1_dm->nrows;
+ MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[tp1], &zt_tm1_sdv, 1.0, 0.0, 'N');
+ MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
+ btp1_sdv.v = bt_dm->M + tp1*btp1_sdv.n;
+ VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
+ //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
+ CopyMatrix0(Pt_tm1_dc->C[tp1], Vt_dc->C[tp1]);
+ MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
+ MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[tp1], Wnzbynz_dm, -1.0);
+ //Done with all W*_dm.
+ MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[tp1], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[tp1], 1.0, 0.0, 'N', 'T');
+ MatrixPlusMatrixUpdate(Pt_tm1_dc->C[tp1], W2nzbynz_dm);
+ //Done with all W*_dm.
+ }
+ zt_tm1_dm->flag = M_GE;
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+ //
+ DestroyVector_lf(wny_dv);
+ DestroyMatrix_lf(Wnzbyny_dm);
+ DestroyMatrix_lf(W2nzbynz_dm);
+ DestroyMatrix_lf(PHtran_tdata_dm);
+ DestroyVector_lf(etdata_dv);
+ DestroyMatrix_lf(Dtdata_dm);
+ DestroyMatrix_lf(Kt_tdata0_dm);
+ DestroyMatrix_lf(Kt_tdata_dm);
+
+ return (loglh_timet);
+}
+
+
+
+
+//-----------------------------------------------------
+//- WARNING: bedore using this function, make sure to call the following functions
+// Only once in creating lwzmodel_ps: Refresh_kalfilms_*(lwzmodel_ps);
+// Everytime when parameters are changed: RefreshEverything(); RefreRunningGensys_allcases(lwzmodel_ps) in particular.
+//-----------------------------------------------------
+double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
+{
+ //st: base-0 grand regime -- deals with the cross-section values at time t.
+ //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
+ // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
+ // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
+ //See (42.3) on p.42 in the SWZII NOTES.
+
+ //-- Output arguments
+ double loglh_timet;
+ //--- Input arguments
+ TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
+ TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
+ //--- Local variables
+ int tbase0;
+ double logdet_Dtdata;
+ //--- Accessible variables
+ int ny = kalfilmsinputs_1stapp_ps->ny;
+ TSdvector etdata_sdv;
+ //=== Work arguments.
+ TSdvector *wny_dv = CreateVector_lf(ny);
+
+
+
+ //--- Critical checking.
+ if (inpt > kalfilmsinputs_1stapp_ps->T)
+ fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): The time exceeds the\n"
+ " data sample size allocated the structure TSkalfilmsinputs_1stapp_tag");
+
+ //--- The following is for safe guard. InitializeKalman_z10_P10() should be called in, say, RefreshEverything().
+ if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
+ if (!InitializeKalman_z10_P10(kalfilmsinputs_1stapp_ps, (TSdmatrix *)NULL, (TSdcell *)NULL))
+ fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): the system is non-stationary when calling"
+ " InitializeKalman_z10_P10(). Please call this function in RefreshEverthing() and"
+ " set the likehood to be -infty for early exit");
+
+ tbase0=inpt-1;
+
+ //------------------- The order matters. Updatekalfilms_1stapp() must be called before Update_et_Dt_1stapp(). -----------------
+ //--- $$$ Critical updating where we MUSt have inpt-1. If inpt, Updatekalfilms_1stapp() will call this function again
+ //--- $$$ because DW function ProbabilityStateConditionalCurrent() need to access this function at time inpt,
+ //--- $$$ which has not computed before Updatekalfilms_1stapp(). Thus, we'll have an infinite loop.
+ Updatekalfilms_1stapp(tbase0, kalfilmsinputs_1stapp_ps, smodel_ps);
+// //--- $$$ Critical updating.
+// Update_et_Dt_1stapp(tbase0, kalfilmsinputs_1stapp_ps);
+// //This function will give Dtdata_d4->F[tbase0], etdata_dc->C[tbase0], and PHtran_tdata_d4->F[tbase0].
+
+
+
+ //======================================================
+ //= Getting the logLH at time tbase0 or time inpt.
+ //======================================================
+ //--- Forming the log conditional likelihood at t.
+ etdata_sdv.n = ny;
+ etdata_sdv.v = etdata_dc->C[tbase0]->M + ny*st;
+ etdata_sdv.flag = V_DEF;
+ if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_d4->F[tbase0]->C[st]))) return (loglh_timet = -NEARINFINITY);
+ bdivA_rgens(wny_dv, &etdata_sdv, '/', Dtdata_d4->F[tbase0]->C[st]);
+ loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, &etdata_sdv);
+ //Done with all w*_dv.
+
+ //===
+ DestroyVector_lf(wny_dv);
+
+ return (loglh_timet);
+}
+//======================================================
+//= Computing z_{1|0} and P_{1|0} for each new parameter values.
+//======================================================
+int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc)
+{
+ //See p.001 and p.004 in LWZ Model II.
+ //Outputs:
+ // return 1: success in initializing; 0: initializing fails, so the likelihood must be set to -infty outside this function.
+ // ztm1_track to track the time up to which Kalman filter have been updated.
+ // z0_dm, zt_tm1_dc->C[0]
+ // P0_dc, Pt_tm1_d4->F[0]
+
+ //--- Output arguments
+ TSdmatrix *z0_0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
+ TSdmatrix *z0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
+ TSdcell *P0_dc = kalfilmsinputs_1stapp_ps->P0_dc; //nz-by-nz-by-nst.
+ //+ Used to get zt_tm1_dc->C[0] and Pt_tm1_d4->F[0] only.
+ TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
+ TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
+ //--- Input arguments
+ TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
+ TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
+ TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
+ //+
+ TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
+ TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
+ TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
+ //--- Local variables
+ int sti;
+ //--- Accessible variables
+ int ny = kalfilmsinputs_1stapp_ps->ny;
+ int nz = kalfilmsinputs_1stapp_ps->nz;
+ int nst = kalfilmsinputs_1stapp_ps->nst;
+ TSdvector z0_sdv, z0_0_sdv, bt_sdv;
+ TSdvector yt_sdv, at_sdv;
+ //--- For the initial conditions: eigenvalue decompositions
+ int ki;
+ int errflag;
+ double eigmax;
+ //===
+ int nz2 = square(nz);
+ TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdvector *wP0_dv = CreateVector_lf(nz2);
+ //
+ TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
+ TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
+
+
+ if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
+ {
+ z0_sdv.n = z0_0_sdv.n = bt_sdv.n = nz;
+ z0_sdv.flag = z0_0_sdv.flag = bt_sdv.flag = V_DEF;
+ at_sdv.n = yt_sdv.n = ny;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+
+
+ //======= Initial condition. =======
+ if (!kalfilmsinputs_1stapp_ps->indxIni)
+ {
+ z0_0_dm->flag = z0_dm->flag = M_GE;
+ for (sti=nst-1; sti>=0; sti--)
+ {
+ if (kalfilmsinputs_1stapp_ps->DiffuseScale) //Diffuse initial conditions are used.
+ {
+ //--- Diffuse condition for z0_dv.
+ z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
+ z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
+ bt_sdv.v = bt_dm->M + bt_sdv.n*sti;
+ InitializeConstantVector_lf(&z0_0_sdv, 0.0);
+ MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
+ VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
+ //--- Diffuse condition for P0_dm.
+ InitializeDiagonalMatrix_lf(Wnzbynz_dm, kalfilmsinputs_1stapp_ps->DiffuseScale); //To be used for DiffuseScale*I(nz)
+ CopyMatrix0(P0_dc->C[sti], Wnzbynz_dm);
+ //Done with W*_dm.
+ }
+ else //Unconditional moments for initial conditions are used.
+ {
+ InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
+ InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
+
+ //=== Eigenanalysis to determine the roots to ensure boundedness.
+ errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti]);
+ if (errflag) fn_DisplayError("kalman.c/InitializeKalman_z10_P10(): eigen decomposition failed");
+ for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
+ evals_abs_dv->flag = V_DEF;
+ eigmax = MaxVector(evals_abs_dv);
+ if (eigmax < (1.0-SQRTEPSILON)) //(1.0+EPSILON))
+ {
+ //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,sti))\b(:,sti);
+ z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
+ z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
+ MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti]);
+ CopySubmatrix2vector(&z0_0_sdv, 0, bt_dm, 0, sti, bt_dm->nrows);
+ bdivA_rgens(&z0_0_sdv, &z0_0_sdv, '\\', Wnzbynz_dm);
+ //- Under the assumption s_0 = s_1 (this is a short-cut).
+ MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
+ VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
+ //Done with Wnzbynz_dm.
+ //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,sti),F(:,:,sti)))\V1(:),n_z,n_z);
+ tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti], Ft_dc->C[sti]);
+ MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
+ CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti], 0, 0, nz2);
+ bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
+ CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
+ //Done with all w*_dv and W*_dm.
+ }
+ else
+ {
+ if (0) //0: no printing.
+ {
+ #if defined (USE_DEBUG_FILE)
+ fprintf(FPTR_DEBUG, "\n-------WARNING: ----------\n");
+ fprintf(FPTR_DEBUG, "\nIn grand regime sti=%d\n", sti);
+ fprintf(FPTR_DEBUG, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
+ " and see p.003 in LWZ Model II");
+ #else
+ fprintf(stdout, "\n-----------------\n");
+ fprintf(stdout, "\nIn grand regime sti=%d\n", sti);
+ fprintf(stdout, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
+ " and see p.003 in LWZ Model II");
+ #endif
+ }
+ //=== See p.000.3 in LWZ Model II.
+ //=== Do NOT use the following option. It turns out that this will often generate explosive conditional liklihood
+ //=== at the end of the sample, because Pt_tm1 shrinks to zero overtime due to the sigularity of
+ //=== the initila condition P_{1|0}.
+ //--- Letting z0_dv = 0.0
+ // z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
+ // InitializeConstantVector_lf(&z0_sdv, 0.0);
+ // //--- Letting P0_dm = V
+ // CopyMatrix0(P0_dc->C[sti], Vt_dc->C[sti]);
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+
+ return (0); //Early exit with kalfilmsinputs_1stapp_ps->ztm1_track continues to be -1.
+ }
+ }
+ }
+ }
+ else
+ {
+ if (!z10_dm) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition z_{1|0}\n"
+ " must be supplied as valid input arguments for when indxIni == 1");
+ else
+ CopyMatrix0(z0_dm, z10_dm);
+
+ if (!P10_dc) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition P_{1|0}\n"
+ " must be supplied as valid input arguments for when indxIni == 1");
+ else
+ CopyCell0(P0_dc, P10_dc);
+ }
+ CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t-1 = 1.
+ CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t-1 = 1.
+
+
+ kalfilmsinputs_1stapp_ps->ztm1_track = 0; //Must reset to 0, meaning initial setting is done and ready for computing LH at t = 1.
+
+ Update_et_Dt_1stapp(0, kalfilmsinputs_1stapp_ps);
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+
+ return (1);
+ }
+ else
+ {
+ fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): calling this function makes sense only if"
+ " kalfilmsinputs_1stapp_ps->ztm1_track is -1. Please check this value.");
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+
+ return (0);
+ }
+}
+//======================================================
+//= Integrating out the lagged regimes in order to
+//= updating zt_tm1 and Pt_tm1 for next perid tp1 through Kim-Nelson filter.
+//= tdata representing base-0 t timing, while inpt represents base-1 t timing.
+//
+//= Purpose: for each inpt, we integrate out grand regimes st
+//= only ONCE to prevent the dimension of updated zt_tm1 and Pt_tm1 through Kim-Nelson filter.
+//======================================================
+int Updatekalfilms_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
+{
+ //Output:
+ // tm1update
+ // z_{t_1+1|t_1}
+ // P_{t_1+1|t_1}
+ //Input:
+ // t-1: base-1 t timing. Thus t-1=inpt-1.
+
+ //--- Local variables
+ int stp1i, sti, t_2, t_2p1;
+ double prob_previous_regimes;
+ //-- Output arguments
+ TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
+ TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
+ //--- Input arguments
+ TSdcell *Gt_dc = kalfilmsinputs_1stapp_ps->Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
+ //+
+ TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
+ TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
+ TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
+ //+
+ TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
+ TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
+ TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
+ //--- Accessible variables
+ int ny = kalfilmsinputs_1stapp_ps->ny;
+ int nz = kalfilmsinputs_1stapp_ps->nz;
+ int nst = kalfilmsinputs_1stapp_ps->nst;
+ int T = kalfilmsinputs_1stapp_ps->T;
+ TSdvector z0_sdv;
+ TSdvector btp1_sdv;
+ TSdvector etdata_sdv;
+ //=== Work arguments.
+ TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
+ //+
+ TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
+ //=== For updating zt_tm1_dm and Pt_tm1.
+ TSdvector *ztp1_t_dv = CreateVector_lf(nz);
+ TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
+ TSdvector *ztp1_dv = CreateVector_lf(nz);
+ TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
+
+
+ //--- Critical checking.
+ if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
+ fn_DisplayError(".../kalman.c/Updatekalfilms_1stapp(): Make sure InitializeKalman_z10_P10() is called in the function RefreshEverthing()");
+
+
+ z0_sdv.n = nz;
+ z0_sdv.flag = V_DEF;
+ btp1_sdv.n = nz;
+ btp1_sdv.flag = V_DEF;
+ //+
+ etdata_sdv.n = ny;
+ etdata_sdv.flag = V_DEF;
+
+ for (t_2=kalfilmsinputs_1stapp_ps->ztm1_track; t_2<t_1; t_2++)
+ {
+ //If t_1 <= ztm1_track, no updating.
+ //If t_1 > ztm1_track, updating z_{t|t-1} and P_{t|t-1} up to t-1 = t_1.
+
+ zt_tm1_dc->C[t_2p1=t_2+1]->flag = M_GE;
+ for (stp1i=nst-1; stp1i>=0; stp1i--)
+ {
+ InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
+ InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
+
+ for (sti=nst-1; sti>=0; sti--)
+ {
+ //=== Updating for next period by integrating out sti..
+ //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
+ //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
+ CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti]);
+ MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i], PHtran_tdata_d4->F[t_2]->C[sti], 1.0, 1.0, 'N', 'N');
+ BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_d4->F[t_2]->C[sti]);
+ //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
+ etdata_sdv.v = etdata_dc->C[t_2]->M + ny*sti;
+ z0_sdv.v = zt_tm1_dc->C[t_2]->M + nz*sti; //sti: regime at time t_2.
+ MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i], &z0_sdv, 1.0, 0.0, 'N');
+ MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, &etdata_sdv, 1.0, 1.0, 'N');
+ btp1_sdv.v = bt_dm->M + stp1i*btp1_sdv.n;
+ VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
+ //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
+ CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
+ MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
+ MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
+ //Done with all W*_dm.
+ MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i], Pt_tm1_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i], 1.0, 0.0, 'N', 'T');
+ MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
+ //Done with all W*_dm.
+
+
+ //--- Integrating out the state at t_2 using
+ //--- P(s_t_2|Y_{t_2}, theta) = ProbabilityStateConditionalCurrent(sti, t_2, smodel_ps);
+ //--- One can also access to P(s_t_2|Y_{t_2}, theta) by using ElementV(smodel_ps->V[t_2],s_{t_2}i),
+ //--- but this access will not call my function logTimetCondLH(), thus no updating for
+ //--- P(s_t_2|Y_{t_2}, and thus leading to incorrect results.
+ //--- ProbabilityStateConditionalCurrent() assumes first data starts at time 1, not 0. Thus
+ //--- we must pass t_2+1 as opposed to t_2.
+ prob_previous_regimes = ProbabilityStateConditionalCurrent(sti, t_2+1, smodel_ps);
+ ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
+ ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
+ Ptp1_dm->flag = M_GE | M_SU | M_SL;
+ //Done with ztp1_t_dv and Ptp1_t_dm.
+ }
+ //--- Filling zt_tm1 and Pt_tm1 for next period.
+ z0_sdv.v = zt_tm1_dc->C[t_2p1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
+ CopyVector0(&z0_sdv, ztp1_dv);
+ CopyMatrix0(Pt_tm1_d4->F[t_2p1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
+ //Done with ztp1_dv, z0_sdv, Ptp1_dm.
+ }
+ //--- $$$ The following is important because it tells ProbabilityStateConditionalCurrent(), which calls
+ //--- $$$ logTimetCondLH_kalfilms_1stapp(), which calls recursively this function again, that there is no
+ //--- $$$ need to update Kalman filter for the period before kalfilmsinputs_1stapp_ps->ztm1_track.
+ kalfilmsinputs_1stapp_ps->ztm1_track = t_2p1; //Means that z_{t_2p1+1|t_2p1} and P_{t_2p1+1|t_2p1} are done.
+
+ //--- $$$ This function must be called after all the above computations are done.
+ Update_et_Dt_1stapp(t_2p1, kalfilmsinputs_1stapp_ps);
+ }
+
+
+ //===
+ DestroyMatrix_lf(Wnzbynz_dm);
+ //
+ DestroyMatrix_lf(Wnzbyny_dm);
+ DestroyMatrix_lf(W2nzbynz_dm);
+ DestroyMatrix_lf(Kt_tdata0_dm);
+ DestroyMatrix_lf(Kt_tdata_dm);
+ //
+ DestroyVector_lf(ztp1_t_dv);
+ DestroyMatrix_lf(Ptp1_t_dm);
+ DestroyVector_lf(ztp1_dv);
+ DestroyMatrix_lf(Ptp1_dm);
+
+ return (kalfilmsinputs_1stapp_ps->ztm1_track);
+}
+//======================================================
+//= Computes etdata and Dtdata for all grand regimes st at tbase0=inpt-1 or dtm1_track
+//= to prevent recomputing this object for different st at given tbase0.
+//======================================================
+static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
+{
+ //Output:
+ // dtm1_track is updated in this function.
+ // PHtran_tdata_d4->F[t-1]
+ // etdata_dc->C[t-1]
+ // Dtdata_d4->F[t-1]
+ //Input:
+ // t_1=inpt-1: base-0 timing for et and Dt before the likelihood at time inpt is computed.
+
+ //--- Local variables
+ int sti, tbase0;
+ //-- Output arguments
+ TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
+ TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
+ TSdcell *yt_tm1_dc = kalfilmsinputs_1stapp_ps->yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0).
+ TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
+ //--- input arguments
+ TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-T.
+ TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-T.
+ //+
+ TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
+ TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
+ TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
+ //--- Accessible variables
+ int ny = kalfilmsinputs_1stapp_ps->ny;
+ int nz = kalfilmsinputs_1stapp_ps->nz;
+ int nst = kalfilmsinputs_1stapp_ps->nst;
+ TSdvector z0_sdv;
+ TSdvector yt_sdv, at_sdv;
+ TSdvector etdata_sdv, yt_tm1_sdv;
+ //=== Work arguments.
+ TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
+
+
+ z0_sdv.n = nz;
+ z0_sdv.flag = V_DEF;
+ at_sdv.n = yt_sdv.n = ny;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+ etdata_sdv.n = yt_tm1_sdv.n = ny;
+ etdata_sdv.flag = yt_tm1_sdv.flag = V_DEF;
+
+ for (tbase0=(kalfilmsinputs_1stapp_ps->dtm1_track+1); tbase0<=t_1; tbase0++)
+ {
+ //Note tbase0<=t_1, NOT tbase0<t_1.
+ //If t_1 < (dtm1_track+1), no updating.
+ //If t_1 >= (dtm1_track+1), updating etdata_dc->C[t-1] and Dtdata_d4->F[t-1] up to t-1=t_1.
+
+ for (sti=nst-1; sti>=0; sti--)
+ {
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti], 1.0, 0.0, 'N', 'T');
+ CopyMatrix0(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4->F[tbase0]->C[sti], PHtran_tdata_dm);
+
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
+ yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
+ at_sdv.v = at_dm->M + sti*at_dm->nrows; //grand regime at time tbase0.
+ z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
+ etdata_sdv.v = etdata_dc->C[tbase0]->M + etdata_sdv.n*sti;
+ yt_tm1_sdv.v = etdata_dc->C[tbase0]->M + yt_tm1_sdv.n*sti;
+ CopyVector0(&yt_tm1_sdv, &at_sdv);
+ MatrixTimesVector(&yt_tm1_sdv, Ht_dc->C[sti], &z0_sdv, 1.0, 1.0, 'N'); //a + H*z_{t|t-1}.
+ VectorMinusVector(&etdata_sdv, &yt_sdv, &yt_tm1_sdv); //y_t - a - H*z_{t|t-1}.
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[sti]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ //Done with z0_sdv.v.
+ ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
+ //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
+ // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
+ // a bad number or a complex number.
+ Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
+ CopyMatrix0(Dtdata_d4->F[tbase0]->C[sti], Dtdata_dm); //Saved to be used for logTimetCondLH_kalfilms_1stapp().
+ }
+
+ //--- $$$ This tracker functions the same way as kalfilmsinputs_1stapp_ps->ztm1_track.
+ kalfilmsinputs_1stapp_ps->dtm1_track = tbase0;
+ }
+
+ //===
+ DestroyMatrix_lf(PHtran_tdata_dm);
+ DestroyMatrix_lf(Dtdata_dm);
+
+ return (kalfilmsinputs_1stapp_ps->dtm1_track);
+}
+
+
+
+
+
+
+//-----------------------------------------------------
+//------------ OLD Code --------------------------
+//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
+//- WARNING: make sure to call the following functions
+// RunningGensys_const7varionly(lwzmodel_ps);
+// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
+//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
+//
+//- IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
+//- the coefficient regime is the 1st state variable, and
+//- the volatility regime is the 2nd state variable.
+//-----------------------------------------------------
+#if defined (NEWVERSIONofDW_SWITCH)
+double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
+{
+ //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
+ // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
+ //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
+ // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
+ // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
+ //See (42.3) on p.42 in the SWZII NOTES.
+
+
+ //--- Local variables
+ int comst_c; //composite (s_tc, s_{t-1}c)
+ int st_c, stm1_c, st_v;
+ int comsti_c; //composite (s_tc, s_{t-1}c)
+ int sti, sti_c, stm1i_c, sti_v;
+ int comstp1i_c; //composite (s_{t+1}c, s_tc)
+ int stp1i, stp1i_c, stp1i_v;
+ int tbase0, tp1;
+ double logdet_Dtdata, loglh_timet;
+ static int record_tbase1_or_inpt_or_tp1 = 0;
+ static int passonce;
+ double prob_previous_regimes;
+ //=== Accessible variables
+ int ny = kalfilmsinputs_ps->ny;
+ int nz = kalfilmsinputs_ps->nz;
+ int nRc = kalfilmsinputs_ps->nRc;
+ int nRstc = kalfilmsinputs_ps->nRstc;
+ int nRv = kalfilmsinputs_ps->nRv;
+ int T = kalfilmsinputs_ps->T;
+ int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
+ int **Index = smodel_ps->sv->index; //Regime-switching states.
+ //smodel_ps->sv->index is for our new code.
+ // For old code (before 9 April 08 and before dsge_switch is created), use smodel_ps->sv->Index;
+ TSdvector z0_sdv;
+ //+ input arguments.
+ TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
+ TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
+ TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
+ //
+ TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
+ TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
+ TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
+ //
+ TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
+ TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
+ //+ Output arguments.
+ TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ //=== Work arguments.
+ int nz2 = square(nz);
+ TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdvector *wP0_dv = CreateVector_lf(nz2);
+ //+
+ TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
+ TSdvector *wny_dv = CreateVector_lf(ny);
+ TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
+ TSdvector *etdata_dv = CreateVector_lf(ny);
+ TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
+ TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
+ //--- For eigenvalue decompositions
+ int ki;
+ int errflag;
+ double eigmax;
+ TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
+ TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
+ //--- For updating zt_tm1_dm and Pt_tm1.
+ TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
+ TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
+ TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
+ TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
+
+
+
+ if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_logTimetLH_kalfilms_1st_approx():\n"
+ " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
+ if (indxIndRegimes && (nRc>1) && (nRv>1))
+ if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
+ " Number of state variables must be coincide with indxIndRegimes");
+
+ tbase0 = (tp1=inpt) - 1;
+
+ z0_sdv.n = z0_dm->nrows;
+ z0_sdv.flag = V_DEF;
+ //
+ at_sdv.n = yt_sdv.n = yt_dm->nrows;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+ btp1_sdv.n = bt_dm->nrows;
+ btp1_sdv.flag = V_DEF;
+
+
+ //======= Initial condition. =======
+ if (tbase0==0)
+ {
+ for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
+ {
+ if (indxIndRegimes)
+ {
+ if (nRc==1) //Volatility.
+ {
+ comsti_c = sti_c = 0;
+ sti_v = sti;
+ }
+ else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
+ {
+ comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
+ sti_v = Index[sti][1]; //volatility state s_tv
+ sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
+ stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
+ }
+ else if ((nRv==1) && (nRc>nRstc))
+ {
+ comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
+ sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
+ stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
+ sti_v = 0;
+ }
+ else if ((nRv==1) && (nRc==nRstc))
+ {
+ comsti_c = sti_c = sti;
+ sti_v = 0;
+ }
+ else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
+ {
+ comsti_c = sti_c = Index[sti][0];
+ sti_v = Index[sti][1];
+ }
+ }
+ else //Syncronized regimes.
+ {
+ if (nRc>nRstc)
+ {
+ comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
+ sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
+ stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
+ sti_v = sti_c;
+ }
+ else
+ comsti_c = sti_c = sti_v = sti;
+ }
+
+
+ if (!kalfilmsinputs_ps->indxIni)
+ {
+ InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
+ InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
+
+ //=== Eigenanalysis to determine the roots to ensure boundedness.
+ errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[comsti_c]);
+ if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
+ for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
+ evals_abs_dv->flag = V_DEF;
+ eigmax = MaxVector(evals_abs_dv);
+ if (eigmax < (1.0+1.0e-14))
+ {
+ //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
+ MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[comsti_c]);
+ z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
+ CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, comsti_c, bt_dm->nrows);
+ bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
+ //Done with Wnzbynz_dm.
+ //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
+ tz_kron(W2nz2bynz2_dm, Ft_dc->C[comsti_c], Ft_dc->C[comsti_c]);
+ MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
+ CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
+//=== ???????? For debugging purpose.
+//if ((inpt<2) && (st==0))
+//{
+// fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
+
+// fprintf(FPTR_DEBUG, "wP0_dv:\n");
+// WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "Vt_dc->C[sti_v=%d]:\n", sti_v);
+// WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
+
+// fflush(FPTR_DEBUG);
+
+//}
+ bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
+ CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
+ //Done with all w*_dv and W*_dm.
+ }
+ else
+ {
+ fprintf(stdout, "\n-----------------\n");
+ fprintf(stdout, "\nIn regime comsti_c=%d and sti_v=%d and at time=%d\n", comsti_c, sti_v, 0);
+ fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
+ " and the initial conditions must be supplied by, say, input arguments");
+ fflush(stdout);
+ }
+ }
+ }
+ z0_dm->flag = M_GE;
+ CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
+ CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
+ }
+
+
+ //======================================================
+ //= Getting the logLH at time tbase0 or time inpt.
+ //======================================================
+ if (indxIndRegimes )
+ {
+ if (nRc==1) //Volatility.
+ {
+ comst_c = st_c = 0;
+ st_v = st;
+ }
+ else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
+ {
+ if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
+ " Number of state variables must be coincide with indxIndRegimes");
+
+ comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
+ st_v = Index[st][1]; //volatility state s_tv
+ st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
+ stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
+ }
+ else if ((nRv==1) && (nRc>nRstc))
+ {
+ comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
+ st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
+ stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
+ st_v = 0;
+ }
+ else if ((nRv==1) && (nRc==nRstc))
+ {
+ comst_c = st_c = st;
+ st_v = 0;
+ }
+ else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
+ {
+ if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
+ " Number of state variables must be coincide with indxIndRegimes");
+
+ comst_c = st_c = Index[st][0];
+ st_v = Index[st][1];
+ }
+ }
+ else //Syncronized regimes
+ {
+ if (nRc>nRstc)
+ {
+ comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
+ st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
+ stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
+ st_v = st_c;
+ }
+ else
+ comst_c = st_c = st_v = st;
+ }
+
+
+ z0_sdv.n = zt_tm1_dc->C[0]->nrows;
+ z0_sdv.flag = V_DEF;
+ //
+ at_sdv.n = yt_sdv.n = yt_dm->nrows;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+
+ //====== Computing the conditional LH at time t. ======
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[comst_c], 1.0, 0.0, 'N', 'T');
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
+ yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
+ at_sdv.v = at_dm->M + comst_c*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
+ z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
+ VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
+ MatrixTimesVector(etdata_dv, Ht_dc->C[comst_c], &z0_sdv, -1.0, 1.0, 'N');
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comst_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
+ //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
+ // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
+ // a bad number or a complex number.
+ Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
+
+
+ //--- Forming the log conditional likelihood at t.
+ if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
+ bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
+//if ((inpt>82) && (inpt<86) )
+//{
+// //Must be declared at the top of this "if" block.
+// int kip1;
+// double tmp_Dtdata;
+// double tmp_expterm;
+
+// fprintf(FPTR_DEBUG, "%%------------------------\n");
+// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
+// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
+
+
+// fprintf(FPTR_DEBUG, "wny_dv:\n");
+// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "etdata_dv:\n");
+// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
+// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %.16e ");
+
+// fflush(FPTR_DEBUG);
+//}
+ loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
+ //Done with all w*_dv.
+
+
+
+
+//=== ???????? For debugging purpose.
+if (inpt==1)
+{
+ double wk1, wk2;
+
+ wk1 = logdet_Dtdata;
+ wk2 = VectorDotVector(wny_dv, etdata_dv);
+ fprintf(FPTR_DEBUG, "logdet_Dtdata = %10.5f\n", wk1);
+ fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f\n", wk2);
+ fprintf(FPTR_DEBUG, "----- etdata_dv: \n");
+ WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
+ fprintf(FPTR_DEBUG, "----- yt_dv: \n");
+ WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
+ fprintf(FPTR_DEBUG, "----- at_dv: \n");
+ WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
+ fprintf(FPTR_DEBUG, "----- z0_dv: \n");
+ WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
+ fprintf(FPTR_DEBUG, "----- Ht_dc->C[comst_c=%d]:\n", comst_c);
+ WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
+
+ fprintf(FPTR_DEBUG, "\n\n");
+
+}
+//
+fprintf(FPTR_DEBUG, " %10.5f\n", loglh_timet);
+fflush(FPTR_DEBUG);
+
+
+//=== ???????? For debugging purpose.
+//fprintf(FPTR_DEBUG, "------------------------\n");
+//fprintf(FPTR_DEBUG, "st=%d and inpt=%d\n", st, inpt);
+//fprintf(FPTR_DEBUG, "loglh_timet = %10.5f\n", loglh_timet);
+//fprintf(FPTR_DEBUG, "&yt_sdv:\n");
+//WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
+////WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
+////fprintf(FPTR_DEBUG, "\n");
+////WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
+//fflush(FPTR_DEBUG);
+
+
+//=== ???????? For debugging purpose.
+//if ((inpt>82) && (inpt<86) )
+//if (inpt<2)
+//{
+// //Must be declared at the top of this "if" block.
+// int kip1;
+// double tmp_Dtdata;
+// double tmp_expterm;
+
+// fprintf(FPTR_DEBUG, "%%------------------------\n");
+// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
+// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
+
+
+// tmp_Dtdata = logdeterminant(Dtdata_dm);
+// tmp_expterm = VectorDotVector(wny_dv, etdata_dv);
+// fprintf(FPTR_DEBUG, "logdeterminant(Dtdata_dm) = %10.5f;\n", tmp_Dtdata);
+// fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f;\n", tmp_expterm);
+// fprintf(FPTR_DEBUG, "wny_dv:\n");
+// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "etdata_dv:\n");
+// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "&yt_sdv:\n");
+// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "&at_sdv:\n");
+// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "&z0_sdv:\n");
+// WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
+// fprintf(FPTR_DEBUG, "Ht_dc->C[comst_c=%d]:\n",comst_c);
+// WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
+// fprintf(FPTR_DEBUG, "Rt_dc->C[st_v=%d]:\n", st_v);
+// WriteMatrix(FPTR_DEBUG, Rt_dc->C[st_v], " %10.5f ");
+// fprintf(FPTR_DEBUG, "Pt_tm1_d4->F[tbase0]->C[st = %d]:\n",st);
+// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[st], " %10.5f ");
+// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
+// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
+
+
+
+
+//// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
+//// fprintf(FPTR_DEBUG, "zt_tm1_dc->C[tbase0]:\n");
+//// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
+//// //WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
+//// //fprintf(FPTR_DEBUG, "\n");
+//// fprintf(FPTR_DEBUG, "bt_dm = [\n");
+//// WriteMatrix(FPTR_DEBUG, bt_dm, " %10.5f ");
+//// fprintf(FPTR_DEBUG, "];\n");
+
+//// fprintf(FPTR_DEBUG, "et:\n");
+//// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
+//// fprintf(FPTR_DEBUG, "yt_dv=[\n");
+//// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
+//// fprintf(FPTR_DEBUG, "]';\n");
+
+//// fprintf(FPTR_DEBUG, "at_dv=[\n");
+//// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
+//// fprintf(FPTR_DEBUG, "]';\n");
+
+
+//// for (ki=0; ki<Ht_dc->ncells; ki++)
+//// {
+//// kip1 = ki+1;
+//// fprintf(FPTR_DEBUG, "Ht_dc(:,:,%d)=[\n", kip1);
+//// WriteMatrix(FPTR_DEBUG, Ht_dc->C[ki], " %10.5f ");
+//// fprintf(FPTR_DEBUG, "];\n");
+//// }
+//// for (ki=0; ki<Ft_dc->ncells; ki++)
+//// {
+//// kip1 = ki+1;
+//// fprintf(FPTR_DEBUG, "Ft_dc(:,:,%d)=[\n", kip1);
+//// WriteMatrix(FPTR_DEBUG, Ft_dc->C[ki], " %10.5f ");
+//// fprintf(FPTR_DEBUG, "];\n");
+//// }
+//// for (ki=0; ki<Vt_dc->ncells; ki++)
+//// {
+//// kip1 = ki+1;
+//// fprintf(FPTR_DEBUG, "Vt_dc(:,:,%d)=[\n", kip1);
+//// WriteMatrix(FPTR_DEBUG, Vt_dc->C[ki], " %10.5f ");
+//// fprintf(FPTR_DEBUG, "];\n");
+//// }
+// fflush(FPTR_DEBUG);
+//}
+
+
+ //======================================================
+ //= Updating zt_tm1 and Pt_tm1 for next perid tp1.
+ //= tdata = tbase0 is base-0 timing.
+ //======================================================
+ if (inpt > record_tbase1_or_inpt_or_tp1) //This condition always satisfies at the 1st period (which is inpt=1).
+ {
+ passonce = 0;
+ record_tbase1_or_inpt_or_tp1 = inpt;
+ }
+ if (!passonce)
+ {
+ for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
+ {
+ if (indxIndRegimes)
+ {
+ if (nRc==1) //Volatility.
+ {
+ comstp1i_c = stp1i_c = 0;
+ stp1i_v = stp1i;
+ }
+ else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
+ {
+ comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
+ stp1i_v = Index[stp1i][1]; //volatility state s_tv
+ stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
+ //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
+ }
+ else if ((nRv==1) && (nRc>nRstc))
+ {
+ comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
+ stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
+ //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
+ stp1i_v = 0;
+ }
+ else if ((nRv==1) && (nRc==nRstc))
+ {
+ comstp1i_c = stp1i_c = stp1i;
+ stp1i_v = 0;
+ }
+ else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
+ {
+ comstp1i_c = stp1i_c = Index[stp1i][0];
+ stp1i_v = Index[stp1i][1];
+ }
+ }
+ else //Syncronized regimes.
+ {
+ if (nRc>nRstc)
+ {
+ comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
+ stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
+ //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
+ stp1i_v = stp1i_c;
+ }
+ else
+ comstp1i_c = stp1i_c = stp1i_v = stp1i;
+ }
+
+
+ InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
+ InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
+
+ for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
+ {
+ if (indxIndRegimes)
+ {
+ if (nRc==1) //Volatility.
+ {
+ comsti_c = sti_c = 0;
+ sti_v = sti;
+ }
+ else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
+ {
+ comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
+ sti_v = Index[sti][1]; //volatility state s_tv
+ sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
+ stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
+ }
+ else if ((nRv==1) && (nRc>nRstc))
+ {
+ comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
+ sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
+ stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
+ sti_v = 0;
+ }
+ else if ((nRv==1) && (nRc==nRstc))
+ {
+ comsti_c = sti_c = sti;
+ sti_v = 0;
+ }
+ else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
+ {
+ comsti_c = sti_c = Index[sti][0];
+ sti_v = Index[sti][1];
+ }
+ }
+ else //Syncronized regimes.
+ {
+ if (nRc>nRstc)
+ {
+ comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
+ sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
+ stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
+ sti_v = sti_c;
+ }
+ else
+ comsti_c = sti_c = sti_v = sti;
+ }
+
+
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[comsti_c], 1.0, 0.0, 'N', 'T');
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
+ yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
+ at_sdv.v = at_dm->M + comsti_c*at_dm->nrows; //comsti_c: coefficient regime at time tbase0.
+ z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
+ VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
+ MatrixTimesVector(etdata_dv, Ht_dc->C[comsti_c], &z0_sdv, -1.0, 1.0, 'N');
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comsti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ //Done with z0_sdv.v.
+ ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
+ //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
+ // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
+ // a bad number or a complex number.
+ Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
+
+
+ //=== Updating for next period by integrating out sti..
+ if (tp1<T)
+ {
+ //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
+
+ //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
+ //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
+ CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
+ MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
+ //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
+ MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
+ MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
+ btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
+ VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
+ //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
+ CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i_v]);
+ MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
+ MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
+ //Done with all W*_dm.
+ MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
+ MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
+ //Done with all W*_dm.
+
+ //--- Integrating out the state at tbase0 using P(s_t|Y_{t-1}, theta) = ElementV(smodel_ps->Z[inpt],s_{inpt}_i).
+ //--- Note tbase0 = inpt-1 because the data in DW code (ElementV) is base-1.
+ //--- Note at this point, we cannot access to P(s_t|Y_t, theta) = ElementV(smodel_ps->V[inpt],s_{inpt}_i)
+ //--- through DW's code. But we can modify my own code to do this later.
+ prob_previous_regimes = ElementV(smodel_ps->Z[inpt],sti);
+ ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
+ ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
+ Ptp1_dm->flag = M_GE | M_SU | M_SL;
+ //Done with ztp1_t_dv and Ptp1_t_dm.
+ }
+ }
+ //--- Filling zt_tm1 and Pt_tm1 for next period
+ if (tp1<T)
+ {
+ z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
+ CopyVector0(&z0_sdv, ztp1_dv);
+ CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
+ //Done with ztp1_dv, z0_sdv, Ptp1_dm.
+ }
+ }
+ if (tp1<T)
+ zt_tm1_dc->C[tp1]->flag = M_GE;
+ }
+
+
+//=== ???????? For debugging purpose.
+//if ((inpt>60) && (inpt<65) ) //if (inpt<5)
+//{
+// int kip1; //Must be declared at the top of this "if" block.
+
+// fprintf(FPTR_DEBUG, "zt_tm1t=[\n");
+// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
+// fprintf(FPTR_DEBUG, "];\n");
+
+// for (ki=0; ki<Pt_tm1_d4->F[tbase0]->ncells; ki++)
+// {
+// kip1 = ki+1;
+// fprintf(FPTR_DEBUG, "Pt_tm1_d4t(:,:,%d)=[\n", kip1);
+// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[ki], " %10.5f ");
+// fprintf(FPTR_DEBUG, "];\n");
+// }
+
+// fflush(FPTR_DEBUG);
+//}
+
+
+//=== ???????? For debugging purpose.
+fprintf(FPTR_DEBUG, " loglh_timet = %10.5f\n", loglh_timet);
+fflush(FPTR_DEBUG);
+
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+ //
+ DestroyVector_lf(wny_dv);
+ DestroyMatrix_lf(Wnzbyny_dm);
+ DestroyMatrix_lf(W2nzbynz_dm);
+ DestroyMatrix_lf(PHtran_tdata_dm);
+ DestroyVector_lf(etdata_dv);
+ DestroyMatrix_lf(Dtdata_dm);
+ DestroyMatrix_lf(Kt_tdata0_dm);
+ DestroyMatrix_lf(Kt_tdata_dm);
+ //
+ DestroyVector_lf(ztp1_t_dv);
+ DestroyMatrix_lf(Ptp1_t_dm);
+ DestroyVector_lf(ztp1_dv);
+ DestroyMatrix_lf(Ptp1_dm);
+
+ return (loglh_timet);
+}
+#undef LOG2PI
+#endif
+
+
+/**
+//----------------------------------------------------------------
+//-- Tested OK, but has not use because tz_Refresh_z_T7P_T_in_kalfilms_1st_approx()
+//-- cannot access to ElementV(smodel_ps->V[tp1],sti) or ElementV(smodel_ps->V[tbase0],sti)
+//-- because no likelihood has been formed at all before this function is called.
+//----------------------------------------------------------------
+#define LOG2PI (1.837877066409345e+000) //log(2*pi)
+//-----------------------------------------------------
+//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
+//- WARNING: make sure to call the following functions
+// RunningGensys_const7varionly(lwzmodel_ps);
+// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
+//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
+//-----------------------------------------------------
+void tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
+{
+ double debug1;
+ //--- Local variables
+ int stp1i, stp1i_c, stp1i_v, sti, sti_c, sti_v, tbase0, tp1;
+ //=== Accessible variables
+ int ny = kalfilmsinputs_ps->ny;
+ int nz = kalfilmsinputs_ps->nz;
+ int nRc = kalfilmsinputs_ps->nRc;
+ int nRv = kalfilmsinputs_ps->nRv;
+ int T = kalfilmsinputs_ps->T;
+ int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
+ TSdvector z0_sdv;
+ //+ input arguments.
+ TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
+ TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
+ TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
+ //
+ TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
+ TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
+ TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
+ //
+ TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
+ TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
+ //+ Output arguments.
+ TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ //=== Work arguments.
+ int nz2 = square(nz);
+ TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
+ TSdvector *wP0_dv = CreateVector_lf(nz2);
+ //+
+ TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
+ TSdvector *wny_dv = CreateVector_lf(ny);
+ TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
+ TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
+ TSdvector *etdata_dv = CreateVector_lf(ny);
+ TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
+ TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
+ TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
+ //--- For eigenvalue decompositions
+ int ki;
+ int errflag;
+ double eigmax;
+ TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
+ TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
+ //--- For updating zt_tm1_dm and Pt_tm1.
+ TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
+ TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
+ TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
+ TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
+
+
+ if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
+ " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
+ if (indxIndRegimes && (nRc>1) && (nRv>1))
+ if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
+ " Number of state variables must be coincide with indxIndRegimes");
+
+
+ z0_sdv.n = z0_dm->nrows;
+ z0_sdv.flag = V_DEF;
+ //
+ at_sdv.n = yt_sdv.n = yt_dm->nrows;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+ btp1_sdv.n = bt_dm->nrows;
+ btp1_sdv.flag = V_DEF;
+
+
+ //======= Initial condition. =======
+ for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
+ {
+ if (indxIndRegimes && (nRc==1))
+ {
+ sti_c = 0;
+ sti_v = sti;
+ }
+ else if (indxIndRegimes && (nRv==1))
+ {
+ sti_c = sti;
+ sti_v = 0;
+ }
+ else if (indxIndRegimes)
+ {
+ sti_c = smodel_ps->sv->Index[sti][0];
+ sti_v = smodel_ps->sv->Index[sti][1];
+ }
+ else
+ {
+ sti_c = sti_v = sti;
+ }
+
+
+ if (!kalfilmsinputs_ps->indxIni)
+ {
+ InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
+ InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
+
+ //=== Eigenanalysis to determine the roots to ensure boundedness.
+ errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti_c]);
+ if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
+ for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
+ evals_abs_dv->flag = V_DEF;
+ eigmax = MaxVector(evals_abs_dv);
+ if (eigmax < (1.0+1.0e-14))
+ {
+ //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
+ MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti_c]);
+ z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
+ CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, sti_c, bt_dm->nrows);
+ bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
+ //Done with Wnzbynz_dm.
+ //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
+ tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti_c], Ft_dc->C[sti_c]);
+ MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
+ CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
+ bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
+ CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
+ //Done with all w*_dv and W*_dm.
+ }
+ else
+ {
+ fprintf(stdout, "\n-----------------\n");
+ fprintf(stdout, "\nIn regime sti_c=%d and sti_v=%d and at time=%d\n", sti_c, sti_v, 0);
+ fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
+ " and the initial conditions must be supplied by, say, input arguments");
+ fflush(stdout);
+ }
+ }
+ }
+ z0_dm->flag = M_GE;
+ CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
+ CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
+
+
+// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[0]:\n");
+// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[0], " %.16e ");
+// fprintf(FPTR_DEBUG, "\n");
+// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[0]->C[0]:\n");
+// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[0]->C[0], " %.16e ");
+
+
+ //============== Updating zt_tm1 and Pt_tm1. ==================
+ for (tbase0=0; tbase0<T; tbase0++ )
+ {
+ //tdata = tbase0 is base-0 timing.
+ tp1 = tbase0 + 1; //Next period.
+
+ for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
+ {
+ if (indxIndRegimes && (nRc==1))
+ {
+ stp1i_c = 0;
+ stp1i_v = stp1i;
+ }
+ else if (indxIndRegimes && (nRv==1))
+ {
+ stp1i_c = stp1i;
+ stp1i_v = 0;
+ }
+ else if (indxIndRegimes)
+ {
+ stp1i_c = smodel_ps->sv->Index[stp1i][0];
+ stp1i_v = smodel_ps->sv->Index[stp1i][1];
+ }
+ else
+ {
+ stp1i_c = stp1i_v = stp1i;
+ }
+
+
+ InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
+ InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
+ for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
+ {
+ if (indxIndRegimes && (nRc==1))
+ {
+ sti_c = 0;
+ sti_v = sti;
+ }
+ else if (indxIndRegimes && (nRv==1))
+ {
+ sti_c = sti;
+ sti_v = 0;
+ }
+ else if (indxIndRegimes)
+ {
+ sti_c = smodel_ps->sv->Index[sti][0];
+ sti_v = smodel_ps->sv->Index[sti][1];
+ }
+ else
+ {
+ sti_c = sti_v = sti;
+ }
+
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti_c], 1.0, 0.0, 'N', 'T');
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
+ yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
+ at_sdv.v = at_dm->M + sti_c*at_dm->nrows; //sti_c: coefficient regime at time tbase0.
+ z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
+ VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
+ MatrixTimesVector(etdata_dv, Ht_dc->C[sti_c], &z0_sdv, -1.0, 1.0, 'N');
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ //Done with z0_sdv.v.
+
+
+ //=== Updating for next period by integrating out sti..
+ if (tp1<T)
+ {
+ //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
+
+ //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
+ //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
+ CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
+ MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+ BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
+ //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
+ MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
+ MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
+ btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
+ VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
+ //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
+ CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
+ MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
+ MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
+ //Done with all W*_dm.
+ MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
+ MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
+ MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
+ //Done with all W*_dm.
+
+ //--- Integrating out the state at tbase0 using P(s_t|Y_t, theta) = ElementV(smodel_ps->V[t+1],s_{t+1}_i).
+ //--- Note because the data in DW code (ElementV) is base-1, t+1 is actually tbase0.
+ debug1 = ElementV(smodel_ps->V[tp1],sti); //?????? Debug.
+ //ScalarTimesVectorUpdate(ztp1_dv, ElementV(smodel_ps->V[tp1],sti), ztp1_t_dv);
+ //ScalarTimesMatrix(Ptp1_dm, ElementV(smodel_ps->V[tp1],sti), Ptp1_t_dm, 1.0);
+ ScalarTimesVectorUpdate(ztp1_dv, 0.5, ztp1_t_dv);
+ ScalarTimesMatrix(Ptp1_dm, 0.5, Ptp1_t_dm, 1.0);
+ Ptp1_dm->flag = M_GE | M_SU | M_SL;
+ //Done with ztp1_t_dv and Ptp1_t_dm.
+ }
+ }
+ //--- Filling zt_tm1 and Pt_tm1 for next period
+ if (tp1<T)
+ {
+ z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
+ CopyVector0(&z0_sdv, ztp1_dv);
+ CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
+ //Done with ztp1_dv, z0_sdv, Ptp1_dm.
+ }
+ }
+ if (tp1<T)
+ zt_tm1_dc->C[tp1]->flag = M_GE;
+
+// fprintf(FPTR_DEBUG, "\n &yt_sdv:\n");
+// WriteMatrix(FPTR_DEBUG, &yt_sdv, " %.16e ");
+// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[tp1]:\n");
+// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tp1], " %.16e ");
+// fprintf(FPTR_DEBUG, "\n");
+// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[tp1]->C[0]:\n");
+// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tp1]->C[0], " %.16e ");
+// fprintf(FPTR_DEBUG, "\n");
+// fflush(FPTR_DEBUG);
+
+
+ }
+
+ //===
+ DestroyVector_dz(evals_dzv);
+ DestroyVector_lf(evals_abs_dv);
+ DestroyMatrix_lf(Wnzbynz_dm);
+ DestroyMatrix_lf(Wnz2bynz2_dm);
+ DestroyMatrix_lf(W2nz2bynz2_dm);
+ DestroyVector_lf(wP0_dv);
+ //
+ DestroyVector_lf(wny_dv);
+ DestroyMatrix_lf(Wnzbyny_dm);
+ DestroyMatrix_lf(W2nzbynz_dm);
+ DestroyMatrix_lf(PHtran_tdata_dm);
+ DestroyVector_lf(etdata_dv);
+ DestroyMatrix_lf(Dtdata_dm);
+ DestroyMatrix_lf(Kt_tdata0_dm);
+ DestroyMatrix_lf(Kt_tdata_dm);
+ //
+ DestroyVector_lf(ztp1_t_dv);
+ DestroyMatrix_lf(Ptp1_t_dm);
+ DestroyVector_lf(ztp1_dv);
+ DestroyMatrix_lf(Ptp1_dm);
+}
+//-----------------------------------------------------
+//- Kalman filter at time t for Markov-switching DSGE model.
+//- WARNING: make sure to call the following functions
+// (1) RunningGensys_const7varionly(lwzmodel_ps);
+// (2) Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
+// (3) tz_Refresh_z_T7P_T_in_kalfilms_1st_approx();
+//- before using kalfilms_timet_1st_approx().
+//-----------------------------------------------------
+double tz_kalfilms_timet_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
+{
+ //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
+ // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
+ //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
+ // The 1st element for S_T is S_T[1] while S_T[0] is s_0. The same for (T+1)-by-1 gbeta_dv and nlcoefs-by-(T+1) galpha_dm.
+ // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
+ //See (42.3) on p.42 in the SWZII NOTES.
+
+
+ //--- Local variables
+ int st_c, st_v, tbase0;
+ double loglh_timet;
+ //--- Accessible variables
+ int ny = kalfilmsinputs_ps->ny;
+ int nz = kalfilmsinputs_ps->nz;
+ int nRc = kalfilmsinputs_ps->nRc;
+ int nRv = kalfilmsinputs_ps->nRv;
+ int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
+ TSdvector z0_sdv;
+ //+ input arguments.
+ TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
+ TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
+ TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
+ TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
+ //+ Output arguments.
+ TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ //=== Work arguments.
+ TSdvector yt_sdv, at_sdv;
+ TSdvector *wny_dv = CreateVector_lf(ny);
+ TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
+ TSdvector *etdata_dv = CreateVector_lf(ny);
+ TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
+
+
+ if (smodel_ps->sv->nstates != zt_tm1_dc->C[0]->ncols) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
+ " Make sure that the column dimension of zt_tm1_dc->C is the same as smodel_ps->sv->nstates");
+
+ tbase0 = inpt - 1; //base-0 time t.
+
+ if (indxIndRegimes && (nRc==1))
+ {
+ st_c = 0;
+ st_v = st;
+ }
+ else if (indxIndRegimes && (nRv==1))
+ {
+ st_c = st;
+ st_v = 0;
+ }
+ else if (indxIndRegimes)
+ {
+ if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
+ " Number of state variables must be coincide with indxIndRegimes");
+ st_c = smodel_ps->sv->Index[st][0];
+ st_v = smodel_ps->sv->Index[st][1];
+ }
+ else
+ {
+ st_c = st_v = st;
+ }
+
+
+ z0_sdv.n = zt_tm1_dc->C[0]->nrows;
+ z0_sdv.flag = V_DEF;
+ //
+ at_sdv.n = yt_sdv.n = yt_dm->nrows;
+ at_sdv.flag = yt_sdv.flag = V_DEF;
+
+ //====== Computing the conditional LH at time t. ======
+ //--- Setup.
+ MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st_c], 1.0, 0.0, 'N', 'T');
+
+ //--- Data.
+ //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
+ yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
+ at_sdv.v = at_dm->M + st_c*at_dm->nrows; //st_c: coefficient regime at time tbase0.
+ z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
+ VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
+ MatrixTimesVector(etdata_dv, Ht_dc->C[st_c], &z0_sdv, -1.0, 1.0, 'N');
+ //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
+ CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
+ MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
+
+ //--- Forming the log conditional likelihood at t.
+ bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
+ loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
+ //Done with all w*_dv.
+
+
+ //===
+ DestroyVector_lf(wny_dv);
+ DestroyMatrix_lf(PHtran_tdata_dm);
+ DestroyVector_lf(etdata_dv);
+ DestroyMatrix_lf(Dtdata_dm);
+
+ return (loglh_timet);
+}
+#undef LOG2PI
+/**/
+
+
+
+
diff --git a/CFiles/kalman.h b/CFiles/kalman.h
old mode 100755
new mode 100644
index df661b5160ecd1be8c733931ddfc94fb29d2f7dd..3733a3007fb5262e0973c974f3d4b22a0e48ee9a
--- a/CFiles/kalman.h
+++ b/CFiles/kalman.h
@@ -1,314 +1,314 @@
-#ifndef __KALMAN_H__
- #define __KALMAN_H__
-
- #include "tzmatlab.h"
- #include "mathlib.h"
- #include "switch.h"
- #include "fn_filesetup.h" //Used to call WriteMatrix(FPTR_DEBUG,....).
-
-
- typedef struct TSkalcvfurw_tag {
- //urw: univariate random walk kalman filter. Desigend specially for the 2006 AER SWZ paper.
-
- //=== Input arguments.
- int indx_tvsigmasq; //0: constant siqmasq in Kalman updating (default);
- //1: Keyensian (project-specific) type of time-varying sigmasq in Kalman updating; See pp.37 and 37a in SWZ Learning NOTES;
- //2: project-specific type;
- //3: another project-specific type.
- double sigmasq; //Variance for the residual eps(t) of the measurement equation.
- int fss; //T: effective sample size (excluding lags).
- int kx; //dimension for x(t).
- TSdmatrix *V_dm; //kx-by-kx. Covariance (symmetric and positive definite) matrix for the residual eta(t) of the transition equation.
- TSdvector *ylhtran_dv; //1-by-T of y(t). The term lh means lelf hand side and tran means transpose.
- TSdmatrix *Xrhtran_dm; //kx-by-T of x(t). The term rh means right hand side and tran means transpose.
- TSdvector *z10_dv; //kx-by-1. Initial condition for prediction: z_{1|0}.
- TSdmatrix *P10_dm; //kx-by-kx symmetric matrix. Initial condition for the variance of the prediction: P_{1|0}.
-
- //=== Output arguments.
- TSdvector *zupdate_dv; //kx-by-1. z_{T+1|T}.
- TSdmatrix *Zpredtran_dm; //kx-by-T matrix of one-step predicted values of z(t). [z_{2|1}, ..., z_{t+1|t}, ..., z_{T+1|T}].
- //Set to NULL (no output) if storeZ = 0;
- TSdcell *Ppred_dc; //T cells and kx-by-kx symmetric and positive definite matrix for each cell. Mean square errors of predicted state variables.
- //{P_{2|1}, ..., P{t+1|t}, ..., P{T+1|T}. Set to NULL (no output if storeV = 0).
- TSdvector *ylhtranpred_dv; //1-by-T one-step prediction of y(t) or ylhtran_dv. Added 03/17/05.
-
- //=== Function itself.
- void (*learning_fnc)(struct TSkalcvfurw_tag *, void *);
- } TSkalcvfurw; //urw: univariate random walk.
- //
- typedef void TFlearninguni(struct TSkalcvfurw_tag *, void *); //For linear rational expectations models.
-
-
- //=== Better version is TSkalfilmsinputs_1stapp_tag. Kalman filter for constant or known-time-varying DSGE models.
- typedef struct TSkalfiltv_tag
- {
- //General (known-time-varying) Kalman filter for DSGE models.
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Not used if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as a initial condition
- // and with z_{t+1|t} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as a initial condition
- // and with P_{t+1|t} as the last element. Thus, we can use it as a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-T.
- TSdcell *Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-T.
- TSdcell *Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv; //nz-by-1;
- TSdmatrix *P0_dm; //nz-by-nz.
-
- //=== Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc; //nz-by-nz-T.
- } TSkalfiltv;
-
-
-
- //=== Inputs for filter for Markov-switching DSGE models at any time t.
- typedef struct TSkalfilmsinputs_1stapp_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on the grand regime s_t that follows a Markov-chain process
- // and is taken as given, and
- // eps(t) = Psi_u(t)*u(t), where Psi_u(t) is n_y-by-n_u and u(t) is n_u-by-1;
- // eta(t) = Psi_e(t)*e(t), where Psi_e(t) is n_z-by-n_e and e(t) is n_e-by-1.
- //
- // Inputs at time t are as follows where nst is number of grand regimes (including lagged regime
- // and coefficients and shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nst matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nst 3-D of Markov-switching matrices in the measurement equation.
- // Psi_u is an n_y-by-n_u-by-nst 3-D Markov-switching matrices.
- // R is an n_y-by-n_y-by-nst 3-D of Markov-switching covariance matrices,
- // E(eps(t) * eps(t)') = Psi_u(t)*Psi_u(t)', for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nst 3-D of Markov-switching E(eta_t * eps_t') = Psi_e(t)*Psi_u(t)'.
- // ------
- // b is an n_z-by-nst matrix of Markov-switching input vectors in the state equation with b(:,st) as an initial condition.
- // (alternatively, with the ergodic weighted b(:,st) as an initial condition).
- // F is an n_z-by-n_z-by-nst 3-D of Markov-switching transition matrices in the state equation with F(:,:,st)
- // as an initial condition (alternatively, with the ergodic weighted F(:,:,st) as an initial condition).
- // Psi_e is an n_z-by-n_e-by-nst 3-D Markov-switching matrices.
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices,
- // E(eta(t)*eta(t)') = Psi_e(t)*Psi_e(t)', for the error in the state equation
- // with V(:,:,st) as an initial condition (alternatively, with the ergodic weighted V(:,:,st) as an initial condition).
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nst matrix of initial condition (Not used if indxIni=0).
- // P0 is an n_z-by-n_z-by-nst 3-D of initial condition (Not used if indxIni=0).
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,st))\b(:,st)
- // vec(P0_0m1) = (I-kron(F(:,:,st),F(:,:,st)))\vec(V(:,:,st))
- // Note that all eigenvalues of the matrix F(:,:,st) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha. Revised, April 2008.
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nu; //number of measurement errors.
- int ne; //number of fundamental errors.
- int nst; //number of grand composite regimes (current and past regimes, coefficient and volatility regimes).
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0,
- //0: using the unconditional momnets for any given regime at time 0 (default when indxDiffuse = 0).
- int indxDiffuse; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- double DiffuseScale; //A large (infinity) number when indxDiffuse = 1.
- int ztm1_track; //t-1 = -1: no initial conditions z_{1|0} and P_{1|0} has been computed yet, but will be using InitializeKalman_z10_P10(),
- //t-1 >= 0:T-1: z_{t|t-1} and P_{t|t-1} are updated up to t-1.
- int dtm1_track; //t-1 = -1: no etdata_dc->C[0] or Dtdata_d4->F[0] has been computed yet.
- //t-1 >= 0:T-1: etdata_dc->C[t-1] and Dtdata_d4->F[t-1] are updated up to t-1.
-
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nst.
- TSdcell *Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Psiut_dc; //ny-by-nu-by-nst. Measurement error coefficient matrix.
- TSdcell *Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nst.
- TSdcell *Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Psiet_dc; //nz-by-ne-by-nst. Impact matrix in the state equation.
- TSdcell *Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_0_dm; //nz-by-nst. z_{0|0}.
- TSdmatrix *z0_dm; //nz-by-nst. z_{1|0}.
- TSdcell *P0_dc; //nz-by-nz-by-nst. P_{1|0}
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nst-by-(T+1), where z_{1|0} is an initial condition (1st element with t-1=0 or t=1 for base-1) and
- // the terminal condition z_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, z_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1), where P_{1|0} is an initial condition (1st element with t-1=0) and
- // the terminal condition P_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, P_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- //+ Will be save for updating likelihood and Kalman filter Updatekalfilms_1stapp(), so save time to recompute these objects again.
- TSdfourth *PHtran_tdata_d4; //nz-by-ny-by-nst-T, P_{t|t-1}*H_t'. Saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc; //ny-by-nst-by-T (with base-0 T), forecast errors e_t in the likelihood.
- TSdcell *yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0). Incorrect now (Used to back out structural shocks).
- TSdfourth *Dtdata_d4; //ny-by-ny-nst-by-T, forecast covariance D_t in the likelihood. Saved for updating Kalman filter Updatekalfilms_1stapp().
- } TSkalfilmsinputs_1stapp;
-
-
- //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- //~~~ OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
- //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- typedef struct TSkalfilmsinputs_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs at time t are as follows where nRc is number of regimes for coefficients
- // nRv is number of regimes for volatility (shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nRc matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nRc 3-D of Markov-switching matrices in the measurement equation.
- // R is an n_y-by-n_y-by-nRv 3-D of Markov-switching covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nRv 3-D of Markov-switching E(eta_t * eps_t').
- // ------
- // b is an n_z-by-nRc matrix of Markov-switching input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-nRc 3-D of Markov-switching transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIndRegimes: 1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nRc*nRv matrix of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // z0 is an n_z-by-nRv matrix of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRc*nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nRc; //number of composite regimes (current and past regimes) for coefficients.
- int nRstc; //number of coefficient regimes at time t.
- int nRv; //number of regimes for volatility (shock variances).
- int indxIndRegimes; //1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nRc.
- TSdcell *Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm; //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- TSdcell *P0_dc; //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- } TSkalfilmsinputs;
-
-
-
-
- //--- Functions for univariate random walk kalman filter.
- TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv); //, int storeZ, int storeV);
- TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps);
- void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps);
-
- //--- New Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp2(int ny, int nz, int nu, int ne, int nst, int T);
- struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp2(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
- struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T);
- struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
- int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc);
- double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-
- //--- OLD Code: Functions for general constant Kalman filter.
- struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T);
- struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- //Used to test tz_logTimetCondLH_kalfiltv(). (Done April 08). double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps);
-
- //--- OLD Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T);
- struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps);
- double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps);
- //IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
- // the coefficient regime is the 1st state variable, and
- // the volatility regime is the 2nd state variable.
-#endif
-
+#ifndef __KALMAN_H__
+ #define __KALMAN_H__
+
+ #include "tzmatlab.h"
+ #include "mathlib.h"
+ #include "dw_switch.h"
+ #include "fn_filesetup.h" //Used to call WriteMatrix(FPTR_DEBUG,....).
+
+
+ typedef struct TSkalcvfurw_tag {
+ //urw: univariate random walk kalman filter. Desigend specially for the 2006 AER SWZ paper.
+
+ //=== Input arguments.
+ int indx_tvsigmasq; //0: constant siqmasq in Kalman updating (default);
+ //1: Keyensian (project-specific) type of time-varying sigmasq in Kalman updating; See pp.37 and 37a in SWZ Learning NOTES;
+ //2: project-specific type;
+ //3: another project-specific type.
+ double sigmasq; //Variance for the residual eps(t) of the measurement equation.
+ int fss; //T: effective sample size (excluding lags).
+ int kx; //dimension for x(t).
+ TSdmatrix *V_dm; //kx-by-kx. Covariance (symmetric and positive definite) matrix for the residual eta(t) of the transition equation.
+ TSdvector *ylhtran_dv; //1-by-T of y(t). The term lh means lelf hand side and tran means transpose.
+ TSdmatrix *Xrhtran_dm; //kx-by-T of x(t). The term rh means right hand side and tran means transpose.
+ TSdvector *z10_dv; //kx-by-1. Initial condition for prediction: z_{1|0}.
+ TSdmatrix *P10_dm; //kx-by-kx symmetric matrix. Initial condition for the variance of the prediction: P_{1|0}.
+
+ //=== Output arguments.
+ TSdvector *zupdate_dv; //kx-by-1. z_{T+1|T}.
+ TSdmatrix *Zpredtran_dm; //kx-by-T matrix of one-step predicted values of z(t). [z_{2|1}, ..., z_{t+1|t}, ..., z_{T+1|T}].
+ //Set to NULL (no output) if storeZ = 0;
+ TSdcell *Ppred_dc; //T cells and kx-by-kx symmetric and positive definite matrix for each cell. Mean square errors of predicted state variables.
+ //{P_{2|1}, ..., P{t+1|t}, ..., P{T+1|T}. Set to NULL (no output if storeV = 0).
+ TSdvector *ylhtranpred_dv; //1-by-T one-step prediction of y(t) or ylhtran_dv. Added 03/17/05.
+
+ //=== Function itself.
+ void (*learning_fnc)(struct TSkalcvfurw_tag *, void *);
+ } TSkalcvfurw; //urw: univariate random walk.
+ //
+ typedef void TFlearninguni(struct TSkalcvfurw_tag *, void *); //For linear rational expectations models.
+
+
+ //=== Better version is TSkalfilmsinputs_1stapp_tag. Kalman filter for constant or known-time-varying DSGE models.
+ typedef struct TSkalfiltv_tag
+ {
+ //General (known-time-varying) Kalman filter for DSGE models.
+ // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
+ // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
+ //
+ // State space model is defined as follows:
+ // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
+ // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
+ // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
+ //
+ // Inputs are as follows:
+ // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
+ // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
+ // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
+ // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
+ // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
+ // ------
+ // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
+ // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
+ // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
+ // ------
+ // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ // 0: using the unconditional mean for any given regime at time 0.
+ // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Not used if indxIni=0.)
+ // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Not used if indxIni=0.)
+ //
+ // Outputs are as follows:
+ // loglh is a value of the log likelihood function of the state-space model
+ // under the assumption that errors are multivariate Gaussian.
+ // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as a initial condition
+ // and with z_{t+1|t} as the last element. Thus, we can use it as a base-1 vector.
+ // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as a initial condition
+ // and with P_{t+1|t} as the last element. Thus, we can use it as a base-1 cell.
+ //
+ // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
+ // z0_0m1 = (I-F(:,:,1))\b(:,1)
+ // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
+ // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
+ //
+ // March 2007, written by Tao Zha
+ // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
+
+ //=== Input arguments.
+ int ny; //number of observables.
+ int nz; //number of state variables.
+ int T; //sample size.
+ int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ //0: using the unconditional mean for any given regime at time 0. (Default value)
+ TSdmatrix *yt_dm; //ny-by-T.
+ TSdmatrix *at_dm; //ny-by-T.
+ TSdcell *Ht_dc; //ny-by-nz-by-T.
+ TSdcell *Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc; //nz-by-ny-by-T. Cross-covariance.
+ //
+ TSdmatrix *bt_dm; //nz-by-T.
+ TSdcell *Ft_dc; //nz-by-nz-by-T.
+ TSdcell *Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
+ //
+ TSdvector *z0_dv; //nz-by-1;
+ TSdmatrix *P0_dm; //nz-by-nz.
+
+ //=== Output arguments.
+ double loglh; //log likelihood.
+ TSdmatrix *zt_tm1_dm; //nz-by-T.
+ TSdcell *Pt_tm1_dc; //nz-by-nz-T.
+ } TSkalfiltv;
+
+
+
+ //=== Inputs for filter for Markov-switching DSGE models at any time t.
+ typedef struct TSkalfilmsinputs_1stapp_tag
+ {
+ //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
+ // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
+ // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
+ //
+ // State space model is defined as follows:
+ // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
+ // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
+ // where a(t), H(t), b(t), and F(t) depend on the grand regime s_t that follows a Markov-chain process
+ // and is taken as given, and
+ // eps(t) = Psi_u(t)*u(t), where Psi_u(t) is n_y-by-n_u and u(t) is n_u-by-1;
+ // eta(t) = Psi_e(t)*e(t), where Psi_e(t) is n_z-by-n_e and e(t) is n_e-by-1.
+ //
+ // Inputs at time t are as follows where nst is number of grand regimes (including lagged regime
+ // and coefficients and shock variances):
+ // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
+ // a is an n_y-by-nst matrix of Markov-switching input vectors in the measurement equation.
+ // H is an n_y-by-n_z-by-nst 3-D of Markov-switching matrices in the measurement equation.
+ // Psi_u is an n_y-by-n_u-by-nst 3-D Markov-switching matrices.
+ // R is an n_y-by-n_y-by-nst 3-D of Markov-switching covariance matrices,
+ // E(eps(t) * eps(t)') = Psi_u(t)*Psi_u(t)', for the error in the measurement equation.
+ // G is an n_z-by-n_y-by-nst 3-D of Markov-switching E(eta_t * eps_t') = Psi_e(t)*Psi_u(t)'.
+ // ------
+ // b is an n_z-by-nst matrix of Markov-switching input vectors in the state equation with b(:,st) as an initial condition.
+ // (alternatively, with the ergodic weighted b(:,st) as an initial condition).
+ // F is an n_z-by-n_z-by-nst 3-D of Markov-switching transition matrices in the state equation with F(:,:,st)
+ // as an initial condition (alternatively, with the ergodic weighted F(:,:,st) as an initial condition).
+ // Psi_e is an n_z-by-n_e-by-nst 3-D Markov-switching matrices.
+ // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices,
+ // E(eta(t)*eta(t)') = Psi_e(t)*Psi_e(t)', for the error in the state equation
+ // with V(:,:,st) as an initial condition (alternatively, with the ergodic weighted V(:,:,st) as an initial condition).
+ // ------
+ // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ // 0: using the unconditional mean for any given regime at time 0.
+ // z0 is an n_z-by-nst matrix of initial condition (Not used if indxIni=0).
+ // P0 is an n_z-by-n_z-by-nst 3-D of initial condition (Not used if indxIni=0).
+ //
+ // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
+ // z0_0m1 = (I-F(:,:,st))\b(:,st)
+ // vec(P0_0m1) = (I-kron(F(:,:,st),F(:,:,st)))\vec(V(:,:,st))
+ // Note that all eigenvalues of the matrix F(:,:,st) are inside the unit circle when the state-space model is bounded (stationary).
+ //
+ // November 2007, written by Tao Zha. Revised, April 2008.
+ // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
+
+ //=== Input arguments.
+ int ny; //number of observables.
+ int nz; //number of state variables.
+ int nu; //number of measurement errors.
+ int ne; //number of fundamental errors.
+ int nst; //number of grand composite regimes (current and past regimes, coefficient and volatility regimes).
+ int T; //sample size.
+ int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0,
+ //0: using the unconditional momnets for any given regime at time 0 (default when indxDiffuse = 0).
+ int indxDiffuse; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
+ //0: using the unconditional moments.
+ double DiffuseScale; //A large (infinity) number when indxDiffuse = 1.
+ int ztm1_track; //t-1 = -1: no initial conditions z_{1|0} and P_{1|0} has been computed yet, but will be using InitializeKalman_z10_P10(),
+ //t-1 >= 0:T-1: z_{t|t-1} and P_{t|t-1} are updated up to t-1.
+ int dtm1_track; //t-1 = -1: no etdata_dc->C[0] or Dtdata_d4->F[0] has been computed yet.
+ //t-1 >= 0:T-1: etdata_dc->C[t-1] and Dtdata_d4->F[t-1] are updated up to t-1.
+
+ TSdmatrix *yt_dm; //ny-by-T.
+ TSdmatrix *at_dm; //ny-by-nst.
+ TSdcell *Ht_dc; //ny-by-nz-by-nst.
+ TSdcell *Psiut_dc; //ny-by-nu-by-nst. Measurement error coefficient matrix.
+ TSdcell *Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
+ //
+ TSdmatrix *bt_dm; //nz-by-nst.
+ TSdcell *Ft_dc; //nz-by-nz-by-nst.
+ TSdcell *Psiet_dc; //nz-by-ne-by-nst. Impact matrix in the state equation.
+ TSdcell *Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
+ //
+ TSdmatrix *z0_0_dm; //nz-by-nst. z_{0|0}.
+ TSdmatrix *z0_dm; //nz-by-nst. z_{1|0}.
+ TSdcell *P0_dc; //nz-by-nz-by-nst. P_{1|0}
+
+
+ //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
+ TSdcell *zt_tm1_dc; //nz-by-nst-by-(T+1), where z_{1|0} is an initial condition (1st element with t-1=0 or t=1 for base-1) and
+ // the terminal condition z_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
+ // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, z_{T+1|T}
+ // has not legal value computed in most applications unless in out-of-sample forecasting problems.
+ TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1), where P_{1|0} is an initial condition (1st element with t-1=0) and
+ // the terminal condition P_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
+ // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, P_{T+1|T}
+ // has not legal value computed in most applications unless in out-of-sample forecasting problems.
+ //+ Will be save for updating likelihood and Kalman filter Updatekalfilms_1stapp(), so save time to recompute these objects again.
+ TSdfourth *PHtran_tdata_d4; //nz-by-ny-by-nst-T, P_{t|t-1}*H_t'. Saved only for updating Kalman filter Updatekalfilms_1stapp().
+ TSdcell *etdata_dc; //ny-by-nst-by-T (with base-0 T), forecast errors e_t in the likelihood.
+ TSdcell *yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0). Incorrect now (Used to back out structural shocks).
+ TSdfourth *Dtdata_d4; //ny-by-ny-nst-by-T, forecast covariance D_t in the likelihood. Saved for updating Kalman filter Updatekalfilms_1stapp().
+ } TSkalfilmsinputs_1stapp;
+
+
+ //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ //~~~ OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
+ //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ typedef struct TSkalfilmsinputs_tag
+ {
+ //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
+ // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
+ // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
+ //
+ // State space model is defined as follows:
+ // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
+ // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
+ // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
+ //
+ // Inputs at time t are as follows where nRc is number of regimes for coefficients
+ // nRv is number of regimes for volatility (shock variances):
+ // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
+ // a is an n_y-by-nRc matrix of Markov-switching input vectors in the measurement equation.
+ // H is an n_y-by-n_z-by-nRc 3-D of Markov-switching matrices in the measurement equation.
+ // R is an n_y-by-n_y-by-nRv 3-D of Markov-switching covariance matrices for the error in the measurement equation.
+ // G is an n_z-by-n_y-by-nRv 3-D of Markov-switching E(eta_t * eps_t').
+ // ------
+ // b is an n_z-by-nRc matrix of Markov-switching input vectors in the state equation with b(:,1) as an initial condition.
+ // F is an n_z-by-n_z-by-nRc 3-D of Markov-switching transition matrices in the state equation with F(:,:,1) as an initial condition.
+ // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
+ // ------
+ // indxIndRegimes: 1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
+ // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ // 0: using the unconditional mean for any given regime at time 0.
+ // z0 is an n_z-by-nRc*nRv matrix of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
+ // z0 is an n_z-by-nRv matrix of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
+ // P0 is an n_z-by-n_z-by-nRc*nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
+ // P0 is an n_z-by-n_z-by-nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
+ //
+ // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
+ // z0_0m1 = (I-F(:,:,1))\b(:,1)
+ // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
+ // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
+ //
+ // November 2007, written by Tao Zha
+ // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
+
+ //=== Input arguments.
+ int ny; //number of observables.
+ int nz; //number of state variables.
+ int nRc; //number of composite regimes (current and past regimes) for coefficients.
+ int nRstc; //number of coefficient regimes at time t.
+ int nRv; //number of regimes for volatility (shock variances).
+ int indxIndRegimes; //1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
+ int T; //sample size.
+ int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
+ //0: using the unconditional mean for any given regime at time 0. (Default value)
+ TSdmatrix *yt_dm; //ny-by-T.
+ TSdmatrix *at_dm; //ny-by-nRc.
+ TSdcell *Ht_dc; //ny-by-nz-by-nRc.
+ TSdcell *Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
+ TSdcell *Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
+ //
+ TSdmatrix *bt_dm; //nz-by-nRc.
+ TSdcell *Ft_dc; //nz-by-nz-by-nRc.
+ TSdcell *Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
+ //
+ TSdmatrix *z0_dm; //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
+ TSdcell *P0_dc; //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
+
+
+ //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
+ TSdcell *zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
+ } TSkalfilmsinputs;
+
+
+
+
+ //--- Functions for univariate random walk kalman filter.
+ TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv); //, int storeZ, int storeV);
+ TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps);
+ void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps);
+
+ //--- New Code: Functions for Markov-switching Kalman filter.
+ struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp2(int ny, int nz, int nu, int ne, int nst, int T);
+ struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp2(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
+ struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T);
+ struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
+ int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc);
+ double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
+
+
+
+ //--- OLD Code: Functions for general constant Kalman filter.
+ struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T);
+ struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
+ //Used to test tz_logTimetCondLH_kalfiltv(). (Done April 08). double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
+ double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps);
+
+ //--- OLD Code: Functions for Markov-switching Kalman filter.
+ struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T);
+ struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps);
+ double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps);
+ //IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
+ // the coefficient regime is the 1st state variable, and
+ // the volatility regime is the 2nd state variable.
+#endif
+
diff --git a/CFiles/kalmanOldWorks.c b/CFiles/kalmanOldWorks.c
deleted file mode 100755
index 0918ed169d82cba2c4a99e8f920645ac955172fc..0000000000000000000000000000000000000000
--- a/CFiles/kalmanOldWorks.c
+++ /dev/null
@@ -1,2634 +0,0 @@
-/*===============================================================================================================
- * Check $$$ for important notes.
- * Check <<>> for updating DW's new switch code or questions for DW.
- *
- * kalcvf_urw(): the Kalman filter forward prediction specialized for only a univariate random walk (urw) process.
- *
- * State space model is defined as follows:
- * z(t+1) = z(t)+eta(t) (state or transition equation)
- * y(t) = x(t)'*z(t)+eps(t) (observation or measurement equation)
- * where for this function, eta and eps must be uncorrelated; y(t) must be 1-by-1. Note that
- * x(t): k-by-1;
- * z(t): k-by-1;
- * eps(t): 1-by-1 and ~ N(0, sigma^2);
- * eta(t): ~ N(0, V) where V is a k-by-k covariance matrix.
- *
- *
- * Written by Tao Zha, May 2004.
- * Revised, May 2008;
-=================================================================================================================*/
-
-/**
-//=== For debugging purpose.
-if (1)
-{
- double t_loglht;
-
- t_loglht = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, " %10.5f\n", t_loglht);
-
- fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
- fprintf(FPTR_DEBUG, "\n wP0_dv:\n");
- WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "\n Vt_dc->C[sti_v=%d]:\n", sti_v);
- WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
- fflush(FPTR_DEBUG);
-}
-/**/
-
-
-#include "kalman.h"
-
-
-static int Updatekalfilms_1stapp(int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv) //, int storeZ, int storeV)
-{
- int _i;
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
- //---
- TSkalcvfurw *kalcvfurw_ps = tzMalloc(1, TSkalcvfurw);
-
-
- kalcvfurw_ps->indx_tvsigmasq = tv;
- kalcvfurw_ps->fss = T;
- kalcvfurw_ps->kx = k;
-
- //===
- kalcvfurw_ps->V_dm = CreateMatrix_lf(k, k);
- kalcvfurw_ps->ylhtran_dv = CreateVector_lf(T);
- kalcvfurw_ps->Xrhtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->z10_dv = CreateVector_lf(k);
- kalcvfurw_ps->P10_dm = CreateMatrix_lf(k, k);
-
- kalcvfurw_ps->zupdate_dv = CreateVector_lf(k);
- kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->ylhtranpred_dv = CreateVector_lf(T);
- //
- rows_iv = CreateVector_int(T);
- cols_iv = CreateVector_int(T);
- for (_i=T-1; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // if (!storeZ) kalcvfurw_ps->Zpredtran_dm = (TSdmatrix *)NULL;
- // else kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- // if (!storeV) kalcvfurw_ps->Ppred_dc = (TSdcell *)NULL;
- // else {
- // rows_iv = CreateVector_int(T);
- // cols_iv = CreateVector_int(T);
- // for (_i=T; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- // kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // }
-
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
- return (kalcvfurw_ps);
-}
-
-TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps)
-{
- if (kalcvfurw_ps) {
- DestroyMatrix_lf(kalcvfurw_ps->V_dm);
- DestroyVector_lf(kalcvfurw_ps->ylhtran_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Xrhtran_dm);
- DestroyVector_lf(kalcvfurw_ps->z10_dv);
- DestroyMatrix_lf(kalcvfurw_ps->P10_dm);
-
- DestroyVector_lf(kalcvfurw_ps->zupdate_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Zpredtran_dm);
- DestroyCell_lf(kalcvfurw_ps->Ppred_dc);
- DestroyVector_lf(kalcvfurw_ps->ylhtranpred_dv);
-
- free(kalcvfurw_ps);
- return ((TSkalcvfurw *)NULL);
- }
- else return (kalcvfurw_ps);
-}
-
-
-void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps)
-{
- //See the notes of SWZ regarding the government's updating of the parameters in their Phillips-curve equation.
- //NOTE: make sure that the value of kalcvfurw_ps->sigmasq and other input values are given.
- int ti;
- double workd, workdenominv;
- //---
- int fss, kx;
- double sigmasq_fix = kalcvfurw_ps->sigmasq;
-// double sigmasq;
- TSdmatrix *V_dm;
- TSdmatrix *Zpredtran_dm;
- TSdcell *Ppred_dc;
- TSdvector *ylhtran_dv;
- TSdmatrix *Xrhtran_dm;
- //===
- TSdvector *workkxby1_dv = NULL; //kx-by-1.
-// TSdvector *work1kxby1_dv = NULL; //kx-by-1.
- TSdmatrix *workkxbykx_dm = NULL; //kx-by-kx symmetric and positive positive.
-// //===
-// TSdvector *zbefore_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vbefore_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
-// TSdvector *zafter_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vafter_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
- //******* WARNING: Some dangerous pointer movement to gain efficiency *******
-// double *yt_p;
-// double *Vbefore_p;
-// double *Vafter_p;
- TSdvector xt_sdv;
- TSdvector zbefore_sdv;
- //TSdmatrix Vbefore_sdm;
- TSdvector zafter_sdv;
- //TSdmatrix Vafter_sdm;
-
-
- if (!kalcvfurw_ps) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps must be created");
- if (!kalcvfurw_ps->V_dm || !kalcvfurw_ps->ylhtran_dv || !kalcvfurw_ps->Xrhtran_dm || !kalcvfurw_ps->z10_dv || !kalcvfurw_ps->P10_dm)
- fn_DisplayError(".../kalcvf_urw(): input arguments kalcvfurw_ps->V_dm, kalcvfurw_ps->ylhtran_dv, kalcvfurw_ps->Xrhtran_dm, kalcvfurw_ps->z10_dv, kalcvfurw_ps->P10_dm must be given legal values");
- if (!(kalcvfurw_ps->P10_dm->flag & (M_SU | M_SL))) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps->P10_dm must be symmetric");
- fss = kalcvfurw_ps->fss;
- kx = kalcvfurw_ps->kx;
- V_dm = kalcvfurw_ps->V_dm;
- Zpredtran_dm = kalcvfurw_ps->Zpredtran_dm;
- Ppred_dc = kalcvfurw_ps->Ppred_dc;
- ylhtran_dv = kalcvfurw_ps->ylhtran_dv;
- Xrhtran_dm = kalcvfurw_ps->Xrhtran_dm;
- //---
- xt_sdv.n = kx;
- xt_sdv.flag = V_DEF;
- zbefore_sdv.n = kx;
- zbefore_sdv.flag = V_DEF;
- zafter_sdv.n = kx;
- zafter_sdv.flag = V_DEF;
-
- //=== Memory allocation.
- workkxby1_dv = CreateVector_lf(kx);
- workkxbykx_dm = CreateMatrix_lf(kx, kx);
-
-
- //------- The first period (ti=0). -------
- zbefore_sdv.v = kalcvfurw_ps->z10_dv->v;
- zafter_sdv.v = Zpredtran_dm->M;
- xt_sdv.v = Xrhtran_dm->M;
- //---
-
- workd = ylhtran_dv->v[0] - (kalcvfurw_ps->ylhtranpred_dv->v[0]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, kalcvfurw_ps->P10_dm, &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
-
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(kalcvfurw_ps->z10_dv->v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t];
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
-
-
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[0], kalcvfurw_ps->P10_dm, workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[0]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
-
- //------- The rest of the periods (ti=1:T-1). -------
- for (ti=1; ti<fss; ti++) {
- //NOTE: ti=0 has been taken care of outside of this loop.
- zbefore_sdv.v = Zpredtran_dm->M + (ti-1)*kx;
- zafter_sdv.v = Zpredtran_dm->M + ti*kx;
- xt_sdv.v = Xrhtran_dm->M + ti*kx;
- //---
- workd = ylhtran_dv->v[ti] - (kalcvfurw_ps->ylhtranpred_dv->v[ti]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, Ppred_dc->C[ti-1], &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(zbefore_sdv.v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[ti], Ppred_dc->C[ti-1], workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[ti]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
- }
- CopyVector0(kalcvfurw_ps->zupdate_dv, &zafter_sdv);
- Zpredtran_dm->flag = M_GE;
- kalcvfurw_ps->ylhtranpred_dv->flag = V_DEF;
-
-// DestroyVector_lf(zbefore_dv);
-// DestroyMatrix_lf(Vbefore_dm);
-// DestroyVector_lf(zafter_dv);
-// DestroyMatrix_lf(Vafter_dm);
-
- DestroyVector_lf(workkxby1_dv);
-// DestroyVector_lf(work1kxby1_dv);
- DestroyMatrix_lf(workkxbykx_dm);
-}
-
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- General constant (known-time-varying) Kalman filter for DSGE models.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T)
-{
- int _i;
- //===
- TSivector *rows_iv = CreateVector_int(T);
- TSivector *cols_iv = CreateVector_int(T);
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfiltv_tag *kalfiltv_ps = tzMalloc(1, struct TSkalfiltv_tag);
-
-
- //--- Default value.
- kalfiltv_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfiltv_ps->ny = ny;
- kalfiltv_ps->nz = nz;
- kalfiltv_ps->T = T;
-
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfiltv_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfiltv_ps->at_dm = CreateMatrix_lf(ny, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = nz;
- }
- rows_iv->flag = cols_iv->flag = V_DEF;
- kalfiltv_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->bt_dm = CreateMatrix_lf(nz, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- kalfiltv_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->z0_dv = CreateVector_lf(nz);
- kalfiltv_ps->P0_dm = CreateMatrix_lf(nz, nz);
-
-
- //---
- kalfiltv_ps->zt_tm1_dm = CreateMatrix_lf(nz, T);
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Pt_tm1_dc = CreateCell_lf(rows_iv, cols_iv);
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfiltv_ps);
-
-}
-//---
-struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- if (kalfiltv_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfiltv_ps->yt_dm);
- DestroyMatrix_lf(kalfiltv_ps->at_dm);
- DestroyCell_lf(kalfiltv_ps->Ht_dc);
- DestroyCell_lf(kalfiltv_ps->Rt_dc);
- DestroyCell_lf(kalfiltv_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfiltv_ps->bt_dm);
- DestroyCell_lf(kalfiltv_ps->Ft_dc);
- DestroyCell_lf(kalfiltv_ps->Vt_dc);
- //---
- DestroyVector_lf(kalfiltv_ps->z0_dv);
- DestroyMatrix_lf(kalfiltv_ps->P0_dm);
- //---
- DestroyMatrix_lf(kalfiltv_ps->zt_tm1_dm);
- DestroyCell_lf(kalfiltv_ps->Pt_tm1_dc);
-
-
- //---
- tzDestroy(kalfiltv_ps); //Must be freed last!
-
- return ((struct TSkalfiltv_tag *)NULL);
- }
- else return (kalfiltv_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //--- Default value.
- kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- kalfilmsinputs_1stapp_ps->tm1_track = -1;
- //t-1 = -1: the initial conditions z_{1|0} and P_{1|0} are yet to be computed using InitializeKalman_z10_P10().
- //t-1 >= 0: z_{t|t-1} and P_{t|t-1} are updated up to t where t-1 = 1 to T.
- //--- Other assignments.
- kalfilmsinputs_1stapp_ps->ny = ny;
- kalfilmsinputs_1stapp_ps->nz = nz;
- kalfilmsinputs_1stapp_ps->nst = nst;
- kalfilmsinputs_1stapp_ps->T = T;
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- //--- For output arguments.
- rows_iv = CreateConstantVector_int(T+1, nz);
- cols_iv = CreateConstantVector_int(T+1, nst);
- kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_1stapp_ps);
-}
-//---
-struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- if (kalfilmsinputs_1stapp_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
- //---
- tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
- }
- else return (kalfilmsinputs_1stapp_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_tag *kalfilmsinputs_ps = tzMalloc(1, struct TSkalfilmsinputs_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //--- Default value.
- kalfilmsinputs_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfilmsinputs_ps->ny = ny;
- kalfilmsinputs_ps->nz = nz;
- kalfilmsinputs_ps->nRc = nRc;
- kalfilmsinputs_ps->nRstc = nRstc;
- kalfilmsinputs_ps->nRv = nRv;
- kalfilmsinputs_ps->indxIndRegimes = indxIndRegimes;
- kalfilmsinputs_ps->T = T;
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_ps->at_dm = CreateMatrix_lf(ny, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, ny);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, ny);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_ps->bt_dm = CreateMatrix_lf(nz, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, nz);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- if (indxIndRegimes)
- {
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRc*nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- //--- For output arguments.
- if (indxIndRegimes)
- {
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRc*nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_ps);
-
-}
-//---
-struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps)
-{
- if (kalfilmsinputs_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->z0_dm);
- DestroyCell_lf(kalfilmsinputs_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_ps->Pt_tm1_d4);
- //---
- tzDestroy(kalfilmsinputs_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_tag *)NULL);
- }
- else return (kalfilmsinputs_ps);
-};
-
-
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//-- Constant-parameters (known-time-varying) Kalman filter
-//-----------------------------------------------------
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //General constant (known-time-varying) Kalman filter for DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- int T = kalfiltv_ps->T;
- int Tp1 = T + 1;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- int indx_badlh = 0; //1: bad likelihood with, say, -infinity of the LH value.
- int tdata, ti;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv; //double loglh_tdata; //logdetDtdata.
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- //--- Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
-
-
-
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "Fatal error: tz_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[0], P0_dm);
-
- //====== See p.002 in LiuWZ. ======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++ )
- {
- //Base-0 timing.
- ti = tdata + 1; //Next period.
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (kalfiltv_ps->loglh = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh += -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //loglh += -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (ti<T)
- {
- //Updating only up to tdata=T-2. The values at ti=T or tdata=T-1 will not be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[ti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + ti*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[ti], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + ti*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[ti], Vt_dc->C[ti]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[ti], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[ti], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[ti], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[ti], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (kalfiltv_ps->loglh = loglh);
-}
-/**
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //This function is used to test tz_logTimetCondLH_kalfiltv().
- int T = kalfiltv_ps->T;
- int tdata;
- double loglh;
-
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++) loglh += tz_logTimetCondLH_kalfiltv(0, tdata+1, kalfiltv_ps);
-
- return (loglh);
-}
-/**/
-//-----------------------------------------------------
-//-- Updating Kalman filter at time t for constant-parameters (or known-time-varying) Kalman filter.
-//-----------------------------------------------------
-double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //st: base-0 grand regime at time t, which is just a dummy for this constant-parameter function in order to use
- // Waggoner's automatic functions.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
- //
- //log LH at time t for constant (known-time-varying) Kalman-filter DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood at time t.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // April 2008, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //--- Output arguments.
- double loglh_timet; //log likelihood at time t.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
- //--- Input arguments.
- int tdata, tp1;
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- int T = kalfiltv_ps->T;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
-
-
- tdata = (tp1=inpt) - 1; //Base-0 time.
-
- //======= Initial condition. =======
- if (tdata==0)
- {
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_logTimetCondLH_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(FPTR_DEBUG, "Fatal error: tz_logTimetCondLH_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and thus the initial conditions must be supplied by, say, input arguments");
- fflush(FPTR_DEBUG);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[tdata], P0_dm);
- }
-
-
- //======= Liklihood at time t (see p.002 in LiuWZ). =======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //======= Updating for the next period. =======
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (tp1<T)
- {
- //Updating only up to tdata=T-2, because the values at tp1=T or tdata=T-1 will NOT be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[tp1], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + tp1*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[tp1], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + tp1*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[tp1], Vt_dc->C[tp1]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[tp1], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[tp1], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[tp1], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[tp1], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (loglh_timet);
-}
-
-
-//-----------------------------------------------------
-//- WARNING: bedore using this function, make sure to call the following functions
-// Only once in creating lwzmodel_ps: Refresh_kalfilms_*(lwzmodel_ps);
-// Everytime when parameters are changed: RefreshEverything(); RefreRunningGensys_allcases(lwzmodel_ps) in particular.
-//-----------------------------------------------------
-double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //st: base-0 grand regime -- deals with the cross-section values at time t.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
- //-- Output arguments
- double loglh_timet;
- //--- input arguments
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_1stapp_ps->Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //+
- TSdmatrix *z0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdcell *P0_dc = kalfilmsinputs_1stapp_ps->P0_dc; //nz-by-nz-by-nst.
- //--- Used to get zt_tm1_dc->C[0] and Pt_tm1_d4->F[0] only.
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Local variables
- int tbase0;
- double logdet_Dtdata;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- int T = kalfilmsinputs_1stapp_ps->T;
- TSdvector z0_sdv;
- TSdvector yt_sdv, at_sdv;
- //=== Work arguments.
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
- //--- Critical checking.
- if (inpt > T) fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): The time exceeds the\n"
- " data sample size allocated the structure TSkalfilmsinputs_1stapp_tag");
-
- //--- $$$ Critical updating where we MUSt have inpt-1. If inpt, Updatekalfilms_1stapp() will call this function again
- //--- $$$ because DW function ProbabilityStateConditionalCurrent() need to access this function at time inpt,
- //--- $$$ which has not computed before Updatekalfilms_1stapp(). Thus, we'll have an infinite loop.
- Updatekalfilms_1stapp(tbase0=inpt-1, kalfilmsinputs_1stapp_ps, smodel_ps);
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + st*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[st], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag |= M_SU | M_SL;
-
-
- //--- Forming the log conditional likelihood at t.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
- //===
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
-
-
- return (loglh_timet);
-}
-//======================================================
-//= Computing z_{1|0} and P_{1|0} for each new parameter values.
-//======================================================
-int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc)
-{
- //See p.001 in LWZ Model II.
- //Outputs:
- // return 1: success in initializing; 0: initializing fails, so the likelihood must be set to -infty outside this function.
- // tm1_track to track the time up to which Kalman filter have been updated.
- // z0_dm, zt_tm1_dc->C[0]
- // P0_dc, Pt_tm1_d4->F[0]
-
- //--- input arguments
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //+
- TSdmatrix *z0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdcell *P0_dc = kalfilmsinputs_1stapp_ps->P0_dc; //nz-by-nz-by-nst.
- //--- Used to get zt_tm1_dc->C[0] and Pt_tm1_d4->F[0] only.
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Local variables
- int sti;
- //--- Accessible variables
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv;
- //--- For the initial conditions: eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- //===
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
-
-
- if (kalfilmsinputs_1stapp_ps->tm1_track < 0)
- {
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- //======= Initial condition. =======
- if (!kalfilmsinputs_1stapp_ps->indxIni)
- {
- z0_dm->flag = M_GE;
- for (sti=nst-1; sti>=0; sti--)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti]);
- if (errflag) fn_DisplayError("kalman.c/InitializeKalman_z10_P10(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < 1.0) //(1.0+EPSILON))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,sti))\b(:,sti);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, sti, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,sti),F(:,:,sti)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti], Ft_dc->C[sti]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- if (0) //0: no printing.
- {
- #if defined (USE_DEBUG_FILE)
- fprintf(FPTR_DEBUG, "\n-------WARNING: ----------\n");
- fprintf(FPTR_DEBUG, "\nIn grand regime sti=%d\n", sti);
- fprintf(FPTR_DEBUG, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #else
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn grand regime sti=%d\n", sti);
- fprintf(stdout, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #endif
- }
- //=== See p.000.3 in LWZ Model II.
- //=== Do NOT use the following option. It turns out that this will often generate explosive conditional liklihood
- //=== at the end of the sample, because Pt_tm1 shrinks to zero overtime due to the sigularity of
- //=== the initila condition P_{1|0}.
- //--- Letting z0_dv = 0.0
- // z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- // InitializeConstantVector_lf(&z0_sdv, 0.0);
- // //--- Letting P0_dm = V
- // CopyMatrix0(P0_dc->C[sti], Vt_dc->C[sti]);
- return (0); //Early exit with kalfilmsinputs_1stapp_ps->tm1_track continues to be -1.
- }
-
-// //====== Getting etdata_dv and Dtdata_dm for the likelihood function.
-// //--- Setup.
-// MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st], 1.0, 0.0, 'N', 'T');
-
-// //--- Data.
-// //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
-// yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
-// at_sdv.v = at_dm->M + st*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
-// z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
-// VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
-// MatrixTimesVector(etdata_dv, Ht_dc->C[st], &z0_sdv, -1.0, 1.0, 'N');
-// //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
-// CopyMatrix0(Dtdata_dm, Rt_dc->C[st]);
-// MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
-// ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
-// //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
-// // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
-// // a bad number or a complex number.
-// Dtdata_dm->flag |= M_SU | M_SL;
-
- }
- }
- else
- {
- if (!z10_dm) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition z_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyMatrix0(z0_dm, z10_dm);
-
- if (!P10_dc) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition P_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyCell0(P0_dc, P10_dc);
- }
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t-1 = 1.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t-1 = 1.
-
- kalfilmsinputs_1stapp_ps->tm1_track = 0; //Must reset to 0, meaning initial setting is done and ready for computing LH at t = 1.
-
- return (1);
- }
- else fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): calling this function makes sense only if"
- " kalfilmsinputs_1stapp_ps->tm1_track is -1. Please check this value.");
-}
-
-
-//======================================================
-//= Integrating out the lagged regimes in order to
-//= updating zt_tm1 and Pt_tm1 for next perid tp1 through Kim-Nelson filter.
-//= tdata representing base-0 t timing, while inpt represents base-1 t timing.
-//
-//= Purpose: for each inpt, we integrate out grand regimes st
-//= only ONCE to prevent the dimension of updated zt_tm1 and Pt_tm1 through Kim-Nelson filter.
-//======================================================
-static int Updatekalfilms_1stapp(int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //Output:
- // tm1update
- // z_{inpt+1|inpt}
- // P_{inpt+1|inpt}
- //Input:
- // inpt: base-1 t timing. Thus inpt=1 is the first period.
-
- //--- Local variables
- int stp1i, sti, tbase0, tbase0p1;
- double prob_previous_regimes;
- //-- Output arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-T.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-T.
- //--- input arguments
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_1stapp_ps->Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- int T = kalfilmsinputs_1stapp_ps->T;
- TSdvector z0_sdv;
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- //=== Work arguments.
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- //+
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //=== For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- //--- The following is for safe guard. InitializeKalman_z10_P10() should be called in, say, RefreshEverthing().
- if (kalfilmsinputs_1stapp_ps->tm1_track < 0)
- if (!InitializeKalman_z10_P10(kalfilmsinputs_1stapp_ps, (TSdmatrix *)NULL, (TSdcell *)NULL))
- fn_DisplayError(".../kalman.c/Updatekalfilms_1stapp(): the system is non-stationary when calling"
- " InitializeKalman_z10_P10(). Please call this function in RefreshEverthing() and"
- " set the likehood to be -infty for early exit");
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = nz;
- btp1_sdv.flag = V_DEF;
-
- for (tbase0=kalfilmsinputs_1stapp_ps->tm1_track; tbase0<inpt; tbase0++)
- {
- //If inpt <= tm1_track, no updating.
- //If inpt > tm1_track, updating z_{t|t-1} and P_{t|t-1} up to t-1 = inpt.
-
- zt_tm1_dc->C[tbase0p1=tbase0+1]->flag = M_GE;
- for (stp1i=nst-1; stp1i>=0; stp1i--)
- {
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti*at_dm->nrows; //grand regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[sti], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //=== Updating for next period by integrating out sti..
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
-
- //--- Integrating out the state at tbase0 using
- //--- P(s_tbase0|Y_{tbase0}, theta) = ProbabilityStateConditionalCurrent(sti, tbase0, smodel_ps);
- //--- One can also access to P(s_tbase0|Y_{tbase0}, theta) by using ElementV(smodel_ps->V[tbase0],s_{tbase0}i),
- //--- but this access will not call my function logTimetCondLH(), thus no updating for
- //--- P(s_tbase0|Y_{tbase0}, and thus leading to incorrect results.
- prob_previous_regimes = ProbabilityStateConditionalCurrent(sti, tbase0, smodel_ps);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period.
- z0_sdv.v = zt_tm1_dc->C[tbase0p1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tbase0p1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
-
- //--- $$$ This following is important because it tells ProbabilityStateConditionalCurrent(), which calls
- //--- $$$ logTimetCondLH_kalfilms_1stapp(), which calls recursively this function again, that there is no
- //--- $$$ need to update Kalman filter for the period before kalfilmsinputs_1stapp_ps->tm1_track.
- kalfilmsinputs_1stapp_ps->tm1_track = tbase0p1; //Means that z_{tbase0p1+1|tbase0p1} and P_{tbase0p1+1|tbase0p1} are done.
- }
- }
-
-
- //===
- DestroyMatrix_lf(Wnzbynz_dm);
- //
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (kalfilmsinputs_1stapp_ps->tm1_track);
-}
-
-
-
-
-//-----------------------------------------------------
-//------------ OLD Code --------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//
-//- IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
-//- the coefficient regime is the 1st state variable, and
-//- the volatility regime is the 2nd state variable.
-//-----------------------------------------------------
-double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int comst_c; //composite (s_tc, s_{t-1}c)
- int st_c, stm1_c, st_v;
- int comsti_c; //composite (s_tc, s_{t-1}c)
- int sti, sti_c, stm1i_c, sti_v;
- int comstp1i_c; //composite (s_{t+1}c, s_tc)
- int stp1i, stp1i_c, stp1i_v;
- int tbase0, tp1;
- double logdet_Dtdata, loglh_timet;
- static int record_tbase1_or_inpt_or_tp1 = 0;
- static int passonce;
- double prob_previous_regimes;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRstc = kalfilmsinputs_ps->nRstc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- int **Index = smodel_ps->sv->index; //Regime-switching states.
- //smodel_ps->sv->index is for our new code.
- // For old code (before 9 April 08 and before dsge_switch is created), use smodel_ps->sv->Index;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_logTimetLH_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- tbase0 = (tp1=inpt) - 1;
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (tbase0==0)
- {
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[comsti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[comsti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, comsti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[comsti_c], Ft_dc->C[comsti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
-//=== ???????? For debugging purpose.
-//if ((inpt<2) && (st==0))
-//{
-// fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
-// fprintf(FPTR_DEBUG, "wP0_dv:\n");
-// WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Vt_dc->C[sti_v=%d]:\n", sti_v);
-// WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
-// fflush(FPTR_DEBUG);
-
-//}
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime comsti_c=%d and sti_v=%d and at time=%d\n", comsti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
- }
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- if (indxIndRegimes )
- {
- if (nRc==1) //Volatility.
- {
- comst_c = st_c = 0;
- st_v = st;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_v = Index[st][1]; //volatility state s_tv
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comst_c = st_c = st;
- st_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = st_c = Index[st][0];
- st_v = Index[st][1];
- }
- }
- else //Syncronized regimes
- {
- if (nRc>nRstc)
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = st_c;
- }
- else
- comst_c = st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[comst_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comst_c*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comst_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comst_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //--- Forming the log conditional likelihood at t.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
-//if ((inpt>82) && (inpt<86) )
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %.16e ");
-
-// fflush(FPTR_DEBUG);
-//}
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
-
-
-//=== ???????? For debugging purpose.
-if (inpt==1)
-{
- double wk1, wk2;
-
- wk1 = logdet_Dtdata;
- wk2 = VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, "logdet_Dtdata = %10.5f\n", wk1);
- fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f\n", wk2);
- fprintf(FPTR_DEBUG, "----- etdata_dv: \n");
- WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- yt_dv: \n");
- WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- at_dv: \n");
- WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- z0_dv: \n");
- WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- Ht_dc->C[comst_c=%d]:\n", comst_c);
- WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-
- fprintf(FPTR_DEBUG, "\n\n");
-
-}
-//
-fprintf(FPTR_DEBUG, " %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//fprintf(FPTR_DEBUG, "------------------------\n");
-//fprintf(FPTR_DEBUG, "st=%d and inpt=%d\n", st, inpt);
-//fprintf(FPTR_DEBUG, "loglh_timet = %10.5f\n", loglh_timet);
-//fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-//WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-////WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-////fprintf(FPTR_DEBUG, "\n");
-////WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>82) && (inpt<86) )
-//if (inpt<2)
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// tmp_Dtdata = logdeterminant(Dtdata_dm);
-// tmp_expterm = VectorDotVector(wny_dv, etdata_dv);
-// fprintf(FPTR_DEBUG, "logdeterminant(Dtdata_dm) = %10.5f;\n", tmp_Dtdata);
-// fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f;\n", tmp_expterm);
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&at_sdv:\n");
-// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&z0_sdv:\n");
-// WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Ht_dc->C[comst_c=%d]:\n",comst_c);
-// WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Rt_dc->C[st_v=%d]:\n", st_v);
-// WriteMatrix(FPTR_DEBUG, Rt_dc->C[st_v], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4->F[tbase0]->C[st = %d]:\n",st);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[st], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-
-
-
-
-//// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "zt_tm1_dc->C[tbase0]:\n");
-//// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-//// //WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-//// //fprintf(FPTR_DEBUG, "\n");
-//// fprintf(FPTR_DEBUG, "bt_dm = [\n");
-//// WriteMatrix(FPTR_DEBUG, bt_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-
-//// fprintf(FPTR_DEBUG, "et:\n");
-//// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "yt_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-//// fprintf(FPTR_DEBUG, "at_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-
-//// for (ki=0; ki<Ht_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ht_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ht_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Ft_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ft_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ft_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Vt_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Vt_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Vt_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-// fflush(FPTR_DEBUG);
-//}
-
-
- //======================================================
- //= Updating zt_tm1 and Pt_tm1 for next perid tp1.
- //= tdata = tbase0 is base-0 timing.
- //======================================================
- if (inpt > record_tbase1_or_inpt_or_tp1) //This condition always satisfies at the 1st period (which is inpt=1).
- {
- passonce = 0;
- record_tbase1_or_inpt_or_tp1 = inpt;
- }
- if (!passonce)
- {
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comstp1i_c = stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_v = Index[stp1i][1]; //volatility state s_tv
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comstp1i_c = stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comstp1i_c = stp1i_c = Index[stp1i][0];
- stp1i_v = Index[stp1i][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = stp1i_c;
- }
- else
- comstp1i_c = stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[comsti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comsti_c*at_dm->nrows; //comsti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comsti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comsti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i_v]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_{t-1}, theta) = ElementV(smodel_ps->Z[inpt],s_{inpt}_i).
- //--- Note tbase0 = inpt-1 because the data in DW code (ElementV) is base-1.
- //--- Note at this point, we cannot access to P(s_t|Y_t, theta) = ElementV(smodel_ps->V[inpt],s_{inpt}_i)
- //--- through DW's code. But we can modify my own code to do this later.
- prob_previous_regimes = ElementV(smodel_ps->Z[inpt],sti);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
- }
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>60) && (inpt<65) ) //if (inpt<5)
-//{
-// int kip1; //Must be declared at the top of this "if" block.
-
-// fprintf(FPTR_DEBUG, "zt_tm1t=[\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-
-// for (ki=0; ki<Pt_tm1_d4->F[tbase0]->ncells; ki++)
-// {
-// kip1 = ki+1;
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4t(:,:,%d)=[\n", kip1);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[ki], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-// }
-
-// fflush(FPTR_DEBUG);
-//}
-
-
-//=== ???????? For debugging purpose.
-fprintf(FPTR_DEBUG, " loglh_timet = %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-
-
-
-/**
-//----------------------------------------------------------------
-//-- Tested OK, but has not use because tz_Refresh_z_T7P_T_in_kalfilms_1st_approx()
-//-- cannot access to ElementV(smodel_ps->V[tp1],sti) or ElementV(smodel_ps->V[tbase0],sti)
-//-- because no likelihood has been formed at all before this function is called.
-//----------------------------------------------------------------
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//-----------------------------------------------------
-void tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- double debug1;
- //--- Local variables
- int stp1i, stp1i_c, stp1i_v, sti, sti_c, sti_v, tbase0, tp1;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, sti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti_c], Ft_dc->C[sti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime sti_c=%d and sti_v=%d and at time=%d\n", sti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
-
-
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[0]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[0]->C[0], " %.16e ");
-
-
- //============== Updating zt_tm1 and Pt_tm1. ==================
- for (tbase0=0; tbase0<T; tbase0++ )
- {
- //tdata = tbase0 is base-0 timing.
- tp1 = tbase0 + 1; //Next period.
-
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if (indxIndRegimes)
- {
- stp1i_c = smodel_ps->sv->Index[stp1i][0];
- stp1i_v = smodel_ps->sv->Index[stp1i][1];
- }
- else
- {
- stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti_c*at_dm->nrows; //sti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[sti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_t, theta) = ElementV(smodel_ps->V[t+1],s_{t+1}_i).
- //--- Note because the data in DW code (ElementV) is base-1, t+1 is actually tbase0.
- debug1 = ElementV(smodel_ps->V[tp1],sti); //?????? Debug.
- //ScalarTimesVectorUpdate(ztp1_dv, ElementV(smodel_ps->V[tp1],sti), ztp1_t_dv);
- //ScalarTimesMatrix(Ptp1_dm, ElementV(smodel_ps->V[tp1],sti), Ptp1_t_dm, 1.0);
- ScalarTimesVectorUpdate(ztp1_dv, 0.5, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, 0.5, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
-
-// fprintf(FPTR_DEBUG, "\n &yt_sdv:\n");
-// WriteMatrix(FPTR_DEBUG, &yt_sdv, " %.16e ");
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[tp1]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tp1], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[tp1]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tp1]->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fflush(FPTR_DEBUG);
-
-
- }
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-}
-//-----------------------------------------------------
-//- Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// (1) RunningGensys_const7varionly(lwzmodel_ps);
-// (2) Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-// (3) tz_Refresh_z_T7P_T_in_kalfilms_1st_approx();
-//- before using kalfilms_timet_1st_approx().
-//-----------------------------------------------------
-double tz_kalfilms_timet_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0. The same for (T+1)-by-1 gbeta_dv and nlcoefs-by-(T+1) galpha_dm.
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int st_c, st_v, tbase0;
- double loglh_timet;
- //--- Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- TSdvector yt_sdv, at_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- if (smodel_ps->sv->nstates != zt_tm1_dc->C[0]->ncols) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Make sure that the column dimension of zt_tm1_dc->C is the same as smodel_ps->sv->nstates");
-
- tbase0 = inpt - 1; //base-0 time t.
-
- if (indxIndRegimes && (nRc==1))
- {
- st_c = 0;
- st_v = st;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- st_c = st;
- st_v = 0;
- }
- else if (indxIndRegimes)
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
- st_c = smodel_ps->sv->Index[st][0];
- st_v = smodel_ps->sv->Index[st][1];
- }
- else
- {
- st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + st_c*at_dm->nrows; //st_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[st_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
-
- //--- Forming the log conditional likelihood at t.
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //===
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-/**/
-
-
-
-
diff --git a/CFiles/kalmanOldWorks.h b/CFiles/kalmanOldWorks.h
deleted file mode 100755
index 17523384085b06a3878ec22e1e9853ae4c213485..0000000000000000000000000000000000000000
--- a/CFiles/kalmanOldWorks.h
+++ /dev/null
@@ -1,290 +0,0 @@
-#ifndef __KALMAN_H__
- #define __KALMAN_H__
-
- #include "tzmatlab.h"
- #include "mathlib.h"
- #include "switch.h"
- #include "fn_filesetup.h" //Used to call WriteMatrix(FPTR_DEBUG,....).
-
-
- typedef struct TSkalcvfurw_tag {
- //urw: univariate random walk kalman filter. Desigend specially for the 2006 AER SWZ paper.
-
- //=== Input arguments.
- int indx_tvsigmasq; //0: constant siqmasq in Kalman updating (default);
- //1: Keyensian (project-specific) type of time-varying sigmasq in Kalman updating; See pp.37 and 37a in SWZ Learning NOTES;
- //2: project-specific type;
- //3: another project-specific type.
- double sigmasq; //Variance for the residual eps(t) of the measurement equation.
- int fss; //T: effective sample size (excluding lags).
- int kx; //dimension for x(t).
- TSdmatrix *V_dm; //kx-by-kx. Covariance (symmetric and positive definite) matrix for the residual eta(t) of the transition equation.
- TSdvector *ylhtran_dv; //1-by-T of y(t). The term lh means lelf hand side and tran means transpose.
- TSdmatrix *Xrhtran_dm; //kx-by-T of x(t). The term rh means right hand side and tran means transpose.
- TSdvector *z10_dv; //kx-by-1. Initial condition for prediction: z_{1|0}.
- TSdmatrix *P10_dm; //kx-by-kx symmetric matrix. Initial condition for the variance of the prediction: P_{1|0}.
-
- //=== Output arguments.
- TSdvector *zupdate_dv; //kx-by-1. z_{T+1|T}.
- TSdmatrix *Zpredtran_dm; //kx-by-T matrix of one-step predicted values of z(t). [z_{2|1}, ..., z_{t+1|t}, ..., z_{T+1|T}].
- //Set to NULL (no output) if storeZ = 0;
- TSdcell *Ppred_dc; //T cells and kx-by-kx symmetric and positive definite matrix for each cell. Mean square errors of predicted state variables.
- //{P_{2|1}, ..., P{t+1|t}, ..., P{T+1|T}. Set to NULL (no output if storeV = 0).
- TSdvector *ylhtranpred_dv; //1-by-T one-step prediction of y(t) or ylhtran_dv. Added 03/17/05.
-
- //=== Function itself.
- void (*learning_fnc)(struct TSkalcvfurw_tag *, void *);
- } TSkalcvfurw; //urw: univariate random walk.
- //
- typedef void TFlearninguni(struct TSkalcvfurw_tag *, void *); //For linear rational expectations models.
-
-
- //=== General Kalman filter for constant or known-time-varying DSGE models.
- typedef struct TSkalfiltv_tag
- {
- //General (known-time-varying) Kalman filter for DSGE models.
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Not used if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as a initial condition
- // and with z_{t+1|t} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as a initial condition
- // and with P_{t+1|t} as the last element. Thus, we can use it as a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-T.
- TSdcell *Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-T.
- TSdcell *Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv; //nz-by-1;
- TSdmatrix *P0_dm; //nz-by-nz.
-
- //=== Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc; //nz-by-nz-T.
-
- } TSkalfiltv;
-
-
-
- //=== Inputs for filter for Markov-switching DSGE models at any time t.
- typedef struct TSkalfilmsinputs_1stapp_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on the grand regime s_t that follows a Markov-chain process
- // and is taken as given.
- //
- // Inputs at time t are as follows where nst is number of grand regimes (including lagged regime
- // and coefficients and shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nst matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nst 3-D of Markov-switching matrices in the measurement equation.
- // R is an n_y-by-n_y-by-nst 3-D of Markov-switching covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nst 3-D of Markov-switching E(eta_t * eps_t').
- // ------
- // b is an n_z-by-nst matrix of Markov-switching input vectors in the state equation with b(:,st) as an initial condition.
- // (alternatively, with the ergodic weighted b(:,st) as an initial condition).
- // F is an n_z-by-n_z-by-nst 3-D of Markov-switching transition matrices in the state equation with F(:,:,st)
- // as an initial condition (alternatively, with the ergodic weighted F(:,:,st) as an initial condition).
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation
- // with V(:,:,st) as an initial condition (alternatively, with the ergodic weighted V(:,:,st) as an initial condition).
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nst matrix of initial condition (Not used if indxIni=0).
- // P0 is an n_z-by-n_z-by-nst 3-D of initial condition (Not used if indxIni=0).
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,st))\b(:,st)
- // vec(P0_0m1) = (I-kron(F(:,:,st),F(:,:,st)))\vec(V(:,:,st))
- // Note that all eigenvalues of the matrix F(:,:,st) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha. Revised, April 2008.
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nst; //number of grand composite regimes (current and past regimes, coefficient and volatility regimes).
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- int tm1_track; //t-1 = -1: the initial conditions z_{1|0} and P_{1|0} are yet to be computed using InitializeKalman_z10_P10().
- //t-1 >= 0: z_{t|t-1} and P_{t|t-1} are updated up to t where t-1 = 1 to T.
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nst.
- TSdcell *Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nst.
- TSdcell *Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm; //nz-by-nst.
- TSdcell *P0_dc; //nz-by-nz-by-nst.
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nst-by-(T+1), where z_{1|0} is an initial condition (1st element with t-1=0) and
- // the terminal condition z_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, z_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1), where P_{1|0} is an initial condition (1st element with t-1=0) and
- // the terminal condition P_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, P_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- } TSkalfilmsinputs_1stapp;
-
-
- //=== OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
- typedef struct TSkalfilmsinputs_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs at time t are as follows where nRc is number of regimes for coefficients
- // nRv is number of regimes for volatility (shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nRc matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nRc 3-D of Markov-switching matrices in the measurement equation.
- // R is an n_y-by-n_y-by-nRv 3-D of Markov-switching covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nRv 3-D of Markov-switching E(eta_t * eps_t').
- // ------
- // b is an n_z-by-nRc matrix of Markov-switching input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-nRc 3-D of Markov-switching transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIndRegimes: 1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nRc*nRv matrix of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // z0 is an n_z-by-nRv matrix of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRc*nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nRc; //number of composite regimes (current and past regimes) for coefficients.
- int nRstc; //number of coefficient regimes at time t.
- int nRv; //number of regimes for volatility (shock variances).
- int indxIndRegimes; //1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nRc.
- TSdcell *Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm; //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- TSdcell *P0_dc; //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
-
- } TSkalfilmsinputs;
-
-
-
-
- //--- Functions for univariate random walk kalman filter.
- TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv); //, int storeZ, int storeV);
- TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps);
- void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps);
-
- //--- New Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T);
- struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
- int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc);
- double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-
- //--- OLD Code: Functions for general constant Kalman filter.
- struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T);
- struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- //Used to test tz_logTimetCondLH_kalfiltv(). (Done April 08). double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps);
-
- //--- OLD Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T);
- struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps);
- double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps);
- //IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
- // the coefficient regime is the 1st state variable, and
- // the volatility regime is the 2nd state variable.
-#endif
-
diff --git a/CFiles/kalmanOldWorks2.c b/CFiles/kalmanOldWorks2.c
deleted file mode 100755
index 3a66f3e2f0b70eb94b987c85cb86926570badac9..0000000000000000000000000000000000000000
--- a/CFiles/kalmanOldWorks2.c
+++ /dev/null
@@ -1,2907 +0,0 @@
-/*===============================================================================================================
- * Check $$$ for important notes.
- * Check <<>> for updating DW's new switch code or questions for DW.
- *
- * kalcvf_urw(): the Kalman filter forward prediction specialized for only a univariate random walk (urw) process.
- *
- * State space model is defined as follows:
- * z(t+1) = z(t)+eta(t) (state or transition equation)
- * y(t) = x(t)'*z(t)+eps(t) (observation or measurement equation)
- * where for this function, eta and eps must be uncorrelated; y(t) must be 1-by-1. Note that
- * x(t): k-by-1;
- * z(t): k-by-1;
- * eps(t): 1-by-1 and ~ N(0, sigma^2);
- * eta(t): ~ N(0, V) where V is a k-by-k covariance matrix.
- *
- *
- * Written by Tao Zha, May 2004.
- * Revised, May 2008;
-=================================================================================================================*/
-
-/**
-//=== For debugging purpose.
-if (1)
-{
- double t_loglht;
-
- t_loglht = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, " %10.5f\n", t_loglht);
-
- fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
- fprintf(FPTR_DEBUG, "\n wP0_dv:\n");
- WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "\n Vt_dc->C[sti_v=%d]:\n", sti_v);
- WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
- fflush(FPTR_DEBUG);
-}
-/**/
-
-
-#include "kalman.h"
-
-
-static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
-static int Updatekalfilms_1stapp(int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv) //, int storeZ, int storeV)
-{
- int _i;
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
- //---
- TSkalcvfurw *kalcvfurw_ps = tzMalloc(1, TSkalcvfurw);
-
-
- kalcvfurw_ps->indx_tvsigmasq = tv;
- kalcvfurw_ps->fss = T;
- kalcvfurw_ps->kx = k;
-
- //===
- kalcvfurw_ps->V_dm = CreateMatrix_lf(k, k);
- kalcvfurw_ps->ylhtran_dv = CreateVector_lf(T);
- kalcvfurw_ps->Xrhtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->z10_dv = CreateVector_lf(k);
- kalcvfurw_ps->P10_dm = CreateMatrix_lf(k, k);
-
- kalcvfurw_ps->zupdate_dv = CreateVector_lf(k);
- kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- kalcvfurw_ps->ylhtranpred_dv = CreateVector_lf(T);
- //
- rows_iv = CreateVector_int(T);
- cols_iv = CreateVector_int(T);
- for (_i=T-1; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // if (!storeZ) kalcvfurw_ps->Zpredtran_dm = (TSdmatrix *)NULL;
- // else kalcvfurw_ps->Zpredtran_dm = CreateMatrix_lf(k, T);
- // if (!storeV) kalcvfurw_ps->Ppred_dc = (TSdcell *)NULL;
- // else {
- // rows_iv = CreateVector_int(T);
- // cols_iv = CreateVector_int(T);
- // for (_i=T; _i>=0; _i--) rows_iv->v[_i] = cols_iv->v[_i] = k;
- // kalcvfurw_ps->Ppred_dc = CreateCell_lf(rows_iv, cols_iv);
- // }
-
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
- return (kalcvfurw_ps);
-}
-
-TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps)
-{
- if (kalcvfurw_ps) {
- DestroyMatrix_lf(kalcvfurw_ps->V_dm);
- DestroyVector_lf(kalcvfurw_ps->ylhtran_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Xrhtran_dm);
- DestroyVector_lf(kalcvfurw_ps->z10_dv);
- DestroyMatrix_lf(kalcvfurw_ps->P10_dm);
-
- DestroyVector_lf(kalcvfurw_ps->zupdate_dv);
- DestroyMatrix_lf(kalcvfurw_ps->Zpredtran_dm);
- DestroyCell_lf(kalcvfurw_ps->Ppred_dc);
- DestroyVector_lf(kalcvfurw_ps->ylhtranpred_dv);
-
- free(kalcvfurw_ps);
- return ((TSkalcvfurw *)NULL);
- }
- else return (kalcvfurw_ps);
-}
-
-
-void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps)
-{
- //See the notes of SWZ regarding the government's updating of the parameters in their Phillips-curve equation.
- //NOTE: make sure that the value of kalcvfurw_ps->sigmasq and other input values are given.
- int ti;
- double workd, workdenominv;
- //---
- int fss, kx;
- double sigmasq_fix = kalcvfurw_ps->sigmasq;
-// double sigmasq;
- TSdmatrix *V_dm;
- TSdmatrix *Zpredtran_dm;
- TSdcell *Ppred_dc;
- TSdvector *ylhtran_dv;
- TSdmatrix *Xrhtran_dm;
- //===
- TSdvector *workkxby1_dv = NULL; //kx-by-1.
-// TSdvector *work1kxby1_dv = NULL; //kx-by-1.
- TSdmatrix *workkxbykx_dm = NULL; //kx-by-kx symmetric and positive positive.
-// //===
-// TSdvector *zbefore_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vbefore_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
-// TSdvector *zafter_dv = CreateVector_lf(kalcvfurw_ps->kx);
-// TSdmatrix *Vafter_dm = CreateMatrix_lf(kalcvfurw_ps->kx, kalcvfurw_ps->kx);
- //******* WARNING: Some dangerous pointer movement to gain efficiency *******
-// double *yt_p;
-// double *Vbefore_p;
-// double *Vafter_p;
- TSdvector xt_sdv;
- TSdvector zbefore_sdv;
- //TSdmatrix Vbefore_sdm;
- TSdvector zafter_sdv;
- //TSdmatrix Vafter_sdm;
-
-
- if (!kalcvfurw_ps) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps must be created");
- if (!kalcvfurw_ps->V_dm || !kalcvfurw_ps->ylhtran_dv || !kalcvfurw_ps->Xrhtran_dm || !kalcvfurw_ps->z10_dv || !kalcvfurw_ps->P10_dm)
- fn_DisplayError(".../kalcvf_urw(): input arguments kalcvfurw_ps->V_dm, kalcvfurw_ps->ylhtran_dv, kalcvfurw_ps->Xrhtran_dm, kalcvfurw_ps->z10_dv, kalcvfurw_ps->P10_dm must be given legal values");
- if (!(kalcvfurw_ps->P10_dm->flag & (M_SU | M_SL))) fn_DisplayError(".../kalcvf_urw(): the input argument kalcvfurw_ps->P10_dm must be symmetric");
- fss = kalcvfurw_ps->fss;
- kx = kalcvfurw_ps->kx;
- V_dm = kalcvfurw_ps->V_dm;
- Zpredtran_dm = kalcvfurw_ps->Zpredtran_dm;
- Ppred_dc = kalcvfurw_ps->Ppred_dc;
- ylhtran_dv = kalcvfurw_ps->ylhtran_dv;
- Xrhtran_dm = kalcvfurw_ps->Xrhtran_dm;
- //---
- xt_sdv.n = kx;
- xt_sdv.flag = V_DEF;
- zbefore_sdv.n = kx;
- zbefore_sdv.flag = V_DEF;
- zafter_sdv.n = kx;
- zafter_sdv.flag = V_DEF;
-
- //=== Memory allocation.
- workkxby1_dv = CreateVector_lf(kx);
- workkxbykx_dm = CreateMatrix_lf(kx, kx);
-
-
- //------- The first period (ti=0). -------
- zbefore_sdv.v = kalcvfurw_ps->z10_dv->v;
- zafter_sdv.v = Zpredtran_dm->M;
- xt_sdv.v = Xrhtran_dm->M;
- //---
-
- workd = ylhtran_dv->v[0] - (kalcvfurw_ps->ylhtranpred_dv->v[0]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, kalcvfurw_ps->P10_dm, &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
-
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(kalcvfurw_ps->z10_dv->v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t];
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
-
-
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[0], kalcvfurw_ps->P10_dm, workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[0]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
-
- //------- The rest of the periods (ti=1:T-1). -------
- for (ti=1; ti<fss; ti++) {
- //NOTE: ti=0 has been taken care of outside of this loop.
- zbefore_sdv.v = Zpredtran_dm->M + (ti-1)*kx;
- zafter_sdv.v = Zpredtran_dm->M + ti*kx;
- xt_sdv.v = Xrhtran_dm->M + ti*kx;
- //---
- workd = ylhtran_dv->v[ti] - (kalcvfurw_ps->ylhtranpred_dv->v[ti]=VectorDotVector(&xt_sdv, &zbefore_sdv)); //y_t - x_t'*z_{t-1}.
- SymmatrixTimesVector(workkxby1_dv, Ppred_dc->C[ti-1], &xt_sdv, 1.0, 0.0); //P_{t|t-1} x_t;
- if (!kalcvfurw_ps->indx_tvsigmasq)
- workdenominv = 1.0/(sigmasq_fix + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else if (kalcvfurw_ps->indx_tvsigmasq == 1) //See pp.37 and 37a in SWZ Learning NOTES.
- workdenominv = 1.0/(sigmasq_fix*square(zbefore_sdv.v[0]) + VectorDotVector(&xt_sdv, workkxby1_dv)); //1/[sigma^2 + x_t' P_{t|t-1} x_t]
- else {
- printf(".../kalman.c/kalcvf_urw(): Have not got time to deal with kalcvfurw_ps->indx_tvsigmasq defined in kalman.h other than 0 or 1");
- exit(EXIT_FAILURE);
- }
- //--- Updating z_{t+1|t}.
- CopyVector0(&zafter_sdv, &zbefore_sdv);
- VectorPlusMinusVectorUpdate(&zafter_sdv, workkxby1_dv, workd*workdenominv); //z_{t+1|t} = z_{t|t-1} + P_{t|t-1} x_t [y_t - x_t'*z_{t-1}] / [sigma^2 + x_t' P_{t|t-1} x_t];
- //--- Updating P_{t+1|t}.
- CopyMatrix0(workkxbykx_dm, V_dm);
- VectorTimesSelf(workkxbykx_dm, workkxby1_dv, -workdenominv, 1.0, (V_dm->flag & M_SU) ? 'U' : 'L');
- // - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- MatrixPlusMatrix(Ppred_dc->C[ti], Ppred_dc->C[ti-1], workkxbykx_dm);
- //P_{t|t-1} - P_{t|t-1}*x_t * xt'*P_{t|t-1} / [sigma^2 + x_t' P_{t|t-1} x_t] + V;
- Ppred_dc->C[ti]->flag = M_GE | M_SU | M_SL; //This is necessary because if P10_dm is initialized as diagonal, it will have M_GE | M_SU | M_SL | M_UT | M_LT,
- // which is no longer true for workkxbykx_dm and therefore gives Ppred_dc->C[0] with M_GE only as a result of MatrixPlusMatrix().
- //Done with all work* arrays.
- }
- CopyVector0(kalcvfurw_ps->zupdate_dv, &zafter_sdv);
- Zpredtran_dm->flag = M_GE;
- kalcvfurw_ps->ylhtranpred_dv->flag = V_DEF;
-
-// DestroyVector_lf(zbefore_dv);
-// DestroyMatrix_lf(Vbefore_dm);
-// DestroyVector_lf(zafter_dv);
-// DestroyMatrix_lf(Vafter_dm);
-
- DestroyVector_lf(workkxby1_dv);
-// DestroyVector_lf(work1kxby1_dv);
- DestroyMatrix_lf(workkxbykx_dm);
-}
-
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- General constant (known-time-varying) Kalman filter for DSGE models.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T)
-{
- int _i;
- //===
- TSivector *rows_iv = CreateVector_int(T);
- TSivector *cols_iv = CreateVector_int(T);
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfiltv_tag *kalfiltv_ps = tzMalloc(1, struct TSkalfiltv_tag);
-
-
- //--- Default value.
- kalfiltv_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfiltv_ps->ny = ny;
- kalfiltv_ps->nz = nz;
- kalfiltv_ps->T = T;
-
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfiltv_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfiltv_ps->at_dm = CreateMatrix_lf(ny, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = nz;
- }
- rows_iv->flag = cols_iv->flag = V_DEF;
- kalfiltv_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = ny;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = ny;
- }
- kalfiltv_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->bt_dm = CreateMatrix_lf(nz, T);
- //
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- kalfiltv_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- //
- kalfiltv_ps->z0_dv = CreateVector_lf(nz);
- kalfiltv_ps->P0_dm = CreateMatrix_lf(nz, nz);
-
-
- //---
- kalfiltv_ps->zt_tm1_dm = CreateMatrix_lf(nz, T);
- for (_i=T-1; _i>=0; _i--)
- {
- rows_iv->v[_i] = nz;
- cols_iv->v[_i] = nz;
- }
- kalfiltv_ps->Pt_tm1_dc = CreateCell_lf(rows_iv, cols_iv);
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfiltv_ps);
-
-}
-//---
-struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- if (kalfiltv_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfiltv_ps->yt_dm);
- DestroyMatrix_lf(kalfiltv_ps->at_dm);
- DestroyCell_lf(kalfiltv_ps->Ht_dc);
- DestroyCell_lf(kalfiltv_ps->Rt_dc);
- DestroyCell_lf(kalfiltv_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfiltv_ps->bt_dm);
- DestroyCell_lf(kalfiltv_ps->Ft_dc);
- DestroyCell_lf(kalfiltv_ps->Vt_dc);
- //---
- DestroyVector_lf(kalfiltv_ps->z0_dv);
- DestroyMatrix_lf(kalfiltv_ps->P0_dm);
- //---
- DestroyMatrix_lf(kalfiltv_ps->zt_tm1_dm);
- DestroyCell_lf(kalfiltv_ps->Pt_tm1_dc);
-
-
- //---
- tzDestroy(kalfiltv_ps); //Must be freed last!
-
- return ((struct TSkalfiltv_tag *)NULL);
- }
- else return (kalfiltv_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- New code: Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp2(int ny, int nz, int nu, int ne, int nst, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //=== Default value.
- kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
- kalfilmsinputs_1stapp_ps->ztm1_track = -1;
- kalfilmsinputs_1stapp_ps->dtm1_track = -1;
-
- //--- Other key assignments.
- kalfilmsinputs_1stapp_ps->ny = ny;
- kalfilmsinputs_1stapp_ps->nz = nz;
- kalfilmsinputs_1stapp_ps->nu = nu;
- kalfilmsinputs_1stapp_ps->ne = ne;
- kalfilmsinputs_1stapp_ps->nst = nst;
- kalfilmsinputs_1stapp_ps->T = T;
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- if (nu)
- {
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nu);
- kalfilmsinputs_1stapp_ps->Psiut_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- kalfilmsinputs_1stapp_ps->Psiut_dc = NULL;
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ne);
- kalfilmsinputs_1stapp_ps->Psiet_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- //--- For output arguments.
- rows_iv = CreateConstantVector_int(T+1, nz);
- cols_iv = CreateConstantVector_int(T+1, nst);
- kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_1stapp_ps);
-}
-//---
-struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp2(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- if (kalfilmsinputs_1stapp_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Psiut_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Psiet_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
- //---
- tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
- }
- else return (kalfilmsinputs_1stapp_ps);
-};
-
-struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps = tzMalloc(1, struct TSkalfilmsinputs_1stapp_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //=== Default value.
- kalfilmsinputs_1stapp_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- kalfilmsinputs_1stapp_ps->indxDiffuse = 1; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- kalfilmsinputs_1stapp_ps->DiffuseScale = 100.0;
- kalfilmsinputs_1stapp_ps->ztm1_track = -1;
- kalfilmsinputs_1stapp_ps->dtm1_track = -1;
-
- //--- Other key assignments.
- kalfilmsinputs_1stapp_ps->ny = ny;
- kalfilmsinputs_1stapp_ps->nz = nz;
- kalfilmsinputs_1stapp_ps->nst = nst;
- kalfilmsinputs_1stapp_ps->T = T;
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_1stapp_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_1stapp_ps->at_dm = CreateMatrix_lf(ny, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->bt_dm = CreateMatrix_lf(nz, nst);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_1stapp_ps->z0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- kalfilmsinputs_1stapp_ps->z0_0_dm = CreateMatrix_lf(nz, nst); //nz-by-nst.
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nst.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- //--- For output arguments.
- rows_iv = CreateConstantVector_int(T+1, nz);
- cols_iv = CreateConstantVector_int(T+1, nst);
- kalfilmsinputs_1stapp_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, nz);
- kalfilmsinputs_1stapp_ps->Pt_tm1_d4 = CreateFourth_lf(T+1, rows_iv, cols_iv); //nz-by-nz-by-nst-by-(T+1).
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nst, nz);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->PHtran_tdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- kalfilmsinputs_1stapp_ps->etdata_dc = CreateCell_lf(rows_iv, cols_iv); //ny-by-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = CreateConstantVector_int(T, ny);
- cols_iv = CreateConstantVector_int(T, nst);
- //
- rows_iv = CreateConstantVector_int(nst, ny);
- cols_iv = CreateConstantVector_int(nst, ny);
- kalfilmsinputs_1stapp_ps->Dtdata_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //ny-by-ny-nst-by-T, used for updating Kalman filter Updatekalfilms_1stapp().
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_1stapp_ps);
-}
-//---
-struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- if (kalfilmsinputs_1stapp_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_dm);
- DestroyMatrix_lf(kalfilmsinputs_1stapp_ps->z0_0_dm);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Pt_tm1_d4);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4);
- DestroyCell_lf(kalfilmsinputs_1stapp_ps->etdata_dc);
- DestroyFourth_lf(kalfilmsinputs_1stapp_ps->Dtdata_d4);
- //---
- tzDestroy(kalfilmsinputs_1stapp_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_1stapp_tag *)NULL);
- }
- else return (kalfilmsinputs_1stapp_ps);
-};
-
-
-//-----------------------------------------------------------------------------------------------------------------------
-//-- OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
-//-----------------------------------------------------------------------------------------------------------------------
-struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T)
-{
- //~~~ Creating the structure and initializing the NULL pointers.
- struct TSkalfilmsinputs_tag *kalfilmsinputs_ps = tzMalloc(1, struct TSkalfilmsinputs_tag);
-
- //===
- TSivector *rows_iv = NULL;
- TSivector *cols_iv = NULL;
-
- //--- Default value.
- kalfilmsinputs_ps->indxIni = 0; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0.
- //--- Other assignments.
- kalfilmsinputs_ps->ny = ny;
- kalfilmsinputs_ps->nz = nz;
- kalfilmsinputs_ps->nRc = nRc;
- kalfilmsinputs_ps->nRstc = nRstc;
- kalfilmsinputs_ps->nRv = nRv;
- kalfilmsinputs_ps->indxIndRegimes = indxIndRegimes;
- kalfilmsinputs_ps->T = T;
-
-
- //--------- Creates memory and assigns values. The order matters.
- kalfilmsinputs_ps->yt_dm = CreateMatrix_lf(ny, T);
- kalfilmsinputs_ps->at_dm = CreateMatrix_lf(ny, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, ny);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ht_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, ny);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Rt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, ny);
- kalfilmsinputs_ps->Gt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- kalfilmsinputs_ps->bt_dm = CreateMatrix_lf(nz, nRc);
- //
- rows_iv = CreateConstantVector_int(nRc, nz);
- cols_iv = CreateConstantVector_int(nRc, nz);
- kalfilmsinputs_ps->Ft_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Vt_dc = CreateCell_lf(rows_iv, cols_iv);
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- if (indxIndRegimes)
- {
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRc*nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- kalfilmsinputs_ps->z0_dm = CreateMatrix_lf(nz, nRv); //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->P0_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- //--- For output arguments.
- if (indxIndRegimes)
- {
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRc*nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRc*nRv, nz);
- cols_iv = CreateConstantVector_int(nRc*nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
- else
- {
- if (nRstc != nRv) fn_DisplayError("kalman.c/CreateTSkalfilmsinputs(): nRstc must equal to nRv when indxIndRegimes==0");
- rows_iv = CreateConstantVector_int(T, nz);
- cols_iv = CreateConstantVector_int(T, nRv);
- kalfilmsinputs_ps->zt_tm1_dc = CreateCell_lf(rows_iv, cols_iv); //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- //
- rows_iv = CreateConstantVector_int(nRv, nz);
- cols_iv = CreateConstantVector_int(nRv, nz);
- kalfilmsinputs_ps->Pt_tm1_d4 = CreateFourth_lf(T, rows_iv, cols_iv); //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- rows_iv = DestroyVector_int(rows_iv);
- cols_iv = DestroyVector_int(cols_iv);
- }
-
-
- //===
- DestroyVector_int(rows_iv);
- DestroyVector_int(cols_iv);
-
- return (kalfilmsinputs_ps);
-
-}
-//---
-struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps)
-{
- if (kalfilmsinputs_ps)
- {
- //=== The order matters!
- DestroyMatrix_lf(kalfilmsinputs_ps->yt_dm);
- DestroyMatrix_lf(kalfilmsinputs_ps->at_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ht_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Rt_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Gt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->bt_dm);
- DestroyCell_lf(kalfilmsinputs_ps->Ft_dc);
- DestroyCell_lf(kalfilmsinputs_ps->Vt_dc);
- //---
- DestroyMatrix_lf(kalfilmsinputs_ps->z0_dm);
- DestroyCell_lf(kalfilmsinputs_ps->P0_dc);
- //---
- DestroyCell_lf(kalfilmsinputs_ps->zt_tm1_dc);
- DestroyFourth_lf(kalfilmsinputs_ps->Pt_tm1_d4);
- //---
- tzDestroy(kalfilmsinputs_ps); //Must be freed last!
-
- return ((struct TSkalfilmsinputs_tag *)NULL);
- }
- else return (kalfilmsinputs_ps);
-};
-
-
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//-- Constant-parameters (known-time-varying) Kalman filter
-//-----------------------------------------------------
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //General constant (known-time-varying) Kalman filter for DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Do not enter if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- int T = kalfiltv_ps->T;
- int Tp1 = T + 1;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- int indx_badlh = 0; //1: bad likelihood with, say, -infinity of the LH value.
- int tdata, ti;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv; //double loglh_tdata; //logdetDtdata.
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- //--- Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
-
-
-
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "Fatal error: tz_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[0], P0_dm);
-
- //====== See p.002 in LiuWZ. ======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++ )
- {
- //Base-0 timing.
- ti = tdata + 1; //Next period.
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (kalfiltv_ps->loglh = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh += -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //loglh += -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (ti<T)
- {
- //Updating only up to tdata=T-2. The values at ti=T or tdata=T-1 will not be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[ti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + ti*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[ti], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + ti*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[ti], Vt_dc->C[ti]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[ti], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[ti], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[ti], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[ti], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (kalfiltv_ps->loglh = loglh);
-}
-/**
-double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //This function is used to test tz_logTimetCondLH_kalfiltv().
- int T = kalfiltv_ps->T;
- int tdata;
- double loglh;
-
- loglh = 0.0;
- for (tdata=0; tdata<T; tdata++) loglh += tz_logTimetCondLH_kalfiltv(0, tdata+1, kalfiltv_ps);
-
- return (loglh);
-}
-/**/
-//-----------------------------------------------------
-//-- Updating Kalman filter at time t for constant-parameters (or known-time-varying) Kalman filter.
-//-----------------------------------------------------
-double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps)
-{
- //st: base-0 grand regime at time t, which is just a dummy for this constant-parameter function in order to use
- // Waggoner's automatic functions.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
- //
- //log LH at time t for constant (known-time-varying) Kalman-filter DSGE models (conditional on all the parameters).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood at time t.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Value to be assigned if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as an initial condition (base-0 first element)
- // and with z_{T|T-1} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as though it were a initial condition
- // and with P_{T|T-1} as the last element. Thus, we can use it as though it were a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // April 2008, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //--- Output arguments.
- double loglh_timet; //log likelihood at time t.
- TSdmatrix *zt_tm1_dm = kalfiltv_ps->zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc = kalfiltv_ps->Pt_tm1_dc; //nz-by-nz-T.
- //--- Input arguments.
- int tdata, tp1;
- TSdvector *z0_dv = kalfiltv_ps->z0_dv; //nz-by-1;
- TSdmatrix *P0_dm = kalfiltv_ps->P0_dm; //nz-by-nz.
- int T = kalfiltv_ps->T;
- int ny = kalfiltv_ps->ny;
- int nz = kalfiltv_ps->nz;
- //--- Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, zt_tm1_sdv, ztp1_t_sdv, btp1_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax, logdet_Dtdata;
- TSdzvector *evals_dzv = NULL;
- TSdvector *evals_abs_dv = NULL; //Absolute eigenvalues.
- //--- Input arguments.
- TSdmatrix *yt_dm = kalfiltv_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfiltv_ps->at_dm; //ny-by-T.
- TSdcell *Ht_dc = kalfiltv_ps->Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc = kalfiltv_ps->Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfiltv_ps->Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfiltv_ps->bt_dm; //nz-by-T.
- TSdcell *Ft_dc = kalfiltv_ps->Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc = kalfiltv_ps->Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
-
-
- tdata = (tp1=inpt) - 1; //Base-0 time.
-
- //======= Initial condition. =======
- if (tdata==0)
- {
- //=== Initializing.
- if (!kalfiltv_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- evals_dzv = CreateVector_dz(nz);
- evals_abs_dv = CreateVector_lf(nz);
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[0]);
- if (errflag) fn_DisplayError("tz_logTimetCondLH_kalfiltv() in kalman.c: eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[0]);
- CopySubmatrix2vector(z0_dv, 0, bt_dm, 0, 0, bt_dm->nrows);
- bdivA_rgens(z0_dv, z0_dv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[0], Ft_dc->C[0]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[0], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dm, 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(FPTR_DEBUG, "Fatal error: tz_logTimetCondLH_kalfiltv() in kalman.c: the system is non-stationary solutions\n"
- " and thus the initial conditions must be supplied by, say, input arguments");
- fflush(FPTR_DEBUG);
- exit( EXIT_FAILURE );
- }
- }
- CopySubvector2matrix(zt_tm1_dm, 0, 0, z0_dv, 0, z0_dv->n);
- CopyMatrix0(Pt_tm1_dc->C[tdata], P0_dm);
- }
-
-
- //======= Liklihood at time t (see p.002 in LiuWZ). =======
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- zt_tm1_sdv.n = ztp1_t_sdv.n = zt_tm1_dm->nrows;
- zt_tm1_sdv.flag = ztp1_t_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_dc->C[tdata], Ht_dc->C[tdata], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata;
- yt_sdv.v = yt_dm->M + tdata*yt_dm->nrows;
- at_sdv.v = at_dm->M + tdata*at_dm->nrows;
- zt_tm1_sdv.v = zt_tm1_dm->M + tdata*zt_tm1_dm->nrows;
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[tdata], &zt_tm1_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tdata);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[tdata]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[tdata], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
- //--- Forming the log likelihood.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //======= Updating for the next period. =======
- //--- Updating zt_tm1_dm and Pt_tm1_dc by ztp1_t_sdv and Pt_tm1_dc->C[ti].
- if (tp1<T)
- {
- //Updating only up to tdata=T-2, because the values at tp1=T or tdata=T-1 will NOT be used in the likelihood function.
-
- //- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata;
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[tdata]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[tp1], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata;
- ztp1_t_sdv.v = zt_tm1_dm->M + tp1*zt_tm1_dm->nrows;
- MatrixTimesVector(&ztp1_t_sdv, Ft_dc->C[tp1], &zt_tm1_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(&ztp1_t_sdv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + tp1*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(&ztp1_t_sdv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Pt_tm1_dc->C[tp1], Vt_dc->C[tp1]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Pt_tm1_dc->C[tp1], Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[tp1], Pt_tm1_dc->C[tdata], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[tp1], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Pt_tm1_dc->C[tp1], W2nzbynz_dm);
- //Done with all W*_dm.
- }
- zt_tm1_dm->flag = M_GE;
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
-
- return (loglh_timet);
-}
-
-
-
-
-//-----------------------------------------------------
-//- WARNING: bedore using this function, make sure to call the following functions
-// Only once in creating lwzmodel_ps: Refresh_kalfilms_*(lwzmodel_ps);
-// Everytime when parameters are changed: RefreshEverything(); RefreRunningGensys_allcases(lwzmodel_ps) in particular.
-//-----------------------------------------------------
-double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //st: base-0 grand regime -- deals with the cross-section values at time t.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
- //-- Output arguments
- double loglh_timet;
- //--- Input arguments
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- Local variables
- int tbase0;
- double logdet_Dtdata;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- TSdvector etdata_sdv;
- //=== Work arguments.
- TSdvector *wny_dv = CreateVector_lf(ny);
-
-
-
- //--- Critical checking.
- if (inpt > kalfilmsinputs_1stapp_ps->T)
- fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): The time exceeds the\n"
- " data sample size allocated the structure TSkalfilmsinputs_1stapp_tag");
-
- //--- The following is for safe guard. InitializeKalman_z10_P10() should be called in, say, RefreshEverything().
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- if (!InitializeKalman_z10_P10(kalfilmsinputs_1stapp_ps, (TSdmatrix *)NULL, (TSdcell *)NULL))
- fn_DisplayError(".../kalman.c/logTimetCondLH_kalfilms_1stapp(): the system is non-stationary when calling"
- " InitializeKalman_z10_P10(). Please call this function in RefreshEverthing() and"
- " set the likehood to be -infty for early exit");
-
- tbase0=inpt-1;
-
- //------------------- The order matters. Updatekalfilms_1stapp() must be called before Update_et_Dt_1stapp(). -----------------
- //--- $$$ Critical updating where we MUSt have inpt-1. If inpt, Updatekalfilms_1stapp() will call this function again
- //--- $$$ because DW function ProbabilityStateConditionalCurrent() need to access this function at time inpt,
- //--- $$$ which has not computed before Updatekalfilms_1stapp(). Thus, we'll have an infinite loop.
- Updatekalfilms_1stapp(tbase0, kalfilmsinputs_1stapp_ps, smodel_ps);
-// //--- $$$ Critical updating.
-// Update_et_Dt_1stapp(tbase0, kalfilmsinputs_1stapp_ps);
-// //This function will give Dtdata_d4->F[tbase0], etdata_dc->C[tbase0], and PHtran_tdata_d4->F[tbase0].
-
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- //--- Forming the log conditional likelihood at t.
- etdata_sdv.n = ny;
- etdata_sdv.v = etdata_dc->C[tbase0]->M + ny*st;
- etdata_sdv.flag = V_DEF;
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_d4->F[tbase0]->C[st]))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, &etdata_sdv, '/', Dtdata_d4->F[tbase0]->C[st]);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, &etdata_sdv);
- //Done with all w*_dv.
-
- //===
- DestroyVector_lf(wny_dv);
-
- return (loglh_timet);
-}
-//======================================================
-//= Computing z_{1|0} and P_{1|0} for each new parameter values.
-//======================================================
-int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc)
-{
- //See p.001 and p.004 in LWZ Model II.
- //Outputs:
- // return 1: success in initializing; 0: initializing fails, so the likelihood must be set to -infty outside this function.
- // ztm1_track to track the time up to which Kalman filter have been updated.
- // z0_dm, zt_tm1_dc->C[0]
- // P0_dc, Pt_tm1_d4->F[0]
-
- //--- Output arguments
- TSdmatrix *z0_0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdmatrix *z0_dm = kalfilmsinputs_1stapp_ps->z0_dm; //nz-by-nst.
- TSdcell *P0_dc = kalfilmsinputs_1stapp_ps->P0_dc; //nz-by-nz-by-nst.
- //+ Used to get zt_tm1_dc->C[0] and Pt_tm1_d4->F[0] only.
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Input arguments
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //--- Local variables
- int sti;
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv, z0_0_sdv, bt_sdv;
- TSdvector yt_sdv, at_sdv;
- //--- For the initial conditions: eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- //===
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
-
-
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- {
- z0_sdv.n = z0_0_sdv.n = bt_sdv.n = nz;
- z0_sdv.flag = z0_0_sdv.flag = bt_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (!kalfilmsinputs_1stapp_ps->indxIni)
- {
- z0_0_dm->flag = z0_dm->flag = M_GE;
- for (sti=nst-1; sti>=0; sti--)
- {
- if (kalfilmsinputs_1stapp_ps->DiffuseScale) //Diffuse initial conditions are used.
- {
- //--- Diffuse condition for z0_dv.
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
- bt_sdv.v = bt_dm->M + bt_sdv.n*sti;
- InitializeConstantVector_lf(&z0_0_sdv, 0.0);
- MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
- VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
- //--- Diffuse condition for P0_dm.
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, kalfilmsinputs_1stapp_ps->DiffuseScale); //To be used for DiffuseScale*I(nz)
- CopyMatrix0(P0_dc->C[sti], Wnzbynz_dm);
- //Done with W*_dm.
- }
- else //Unconditional moments for initial conditions are used.
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti]);
- if (errflag) fn_DisplayError("kalman.c/InitializeKalman_z10_P10(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0-SQRTEPSILON)) //(1.0+EPSILON))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,sti))\b(:,sti);
- z0_0_sdv.v = z0_0_dm->M + z0_0_sdv.n*sti;
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti]);
- CopySubmatrix2vector(&z0_0_sdv, 0, bt_dm, 0, sti, bt_dm->nrows);
- bdivA_rgens(&z0_0_sdv, &z0_0_sdv, '\\', Wnzbynz_dm);
- //- Under the assumption s_0 = s_1 (this is a short-cut).
- MatrixTimesVector(&z0_sdv, Ft_dc->C[sti], &z0_0_sdv, 1.0, 0.0, 'N');
- VectorPlusVector(&z0_sdv, &z0_sdv, &bt_sdv);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,sti),F(:,:,sti)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti], Ft_dc->C[sti]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- if (0) //0: no printing.
- {
- #if defined (USE_DEBUG_FILE)
- fprintf(FPTR_DEBUG, "\n-------WARNING: ----------\n");
- fprintf(FPTR_DEBUG, "\nIn grand regime sti=%d\n", sti);
- fprintf(FPTR_DEBUG, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #else
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn grand regime sti=%d\n", sti);
- fprintf(stdout, ".../kalman.c/InitializeKalman_z10_P10(): the system is non-stationary solutions\n"
- " and see p.003 in LWZ Model II");
- #endif
- }
- //=== See p.000.3 in LWZ Model II.
- //=== Do NOT use the following option. It turns out that this will often generate explosive conditional liklihood
- //=== at the end of the sample, because Pt_tm1 shrinks to zero overtime due to the sigularity of
- //=== the initila condition P_{1|0}.
- //--- Letting z0_dv = 0.0
- // z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- // InitializeConstantVector_lf(&z0_sdv, 0.0);
- // //--- Letting P0_dm = V
- // CopyMatrix0(P0_dc->C[sti], Vt_dc->C[sti]);
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (0); //Early exit with kalfilmsinputs_1stapp_ps->ztm1_track continues to be -1.
- }
- }
- }
- }
- else
- {
- if (!z10_dm) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition z_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyMatrix0(z0_dm, z10_dm);
-
- if (!P10_dc) fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): The initial condition P_{1|0}\n"
- " must be supplied as valid input arguments for when indxIni == 1");
- else
- CopyCell0(P0_dc, P10_dc);
- }
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t-1 = 1.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t-1 = 1.
-
-
- kalfilmsinputs_1stapp_ps->ztm1_track = 0; //Must reset to 0, meaning initial setting is done and ready for computing LH at t = 1.
-
- Update_et_Dt_1stapp(0, kalfilmsinputs_1stapp_ps);
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (1);
- }
- else
- {
- fn_DisplayError(".../kalman.c/InitializeKalman_z10_P10(): calling this function makes sense only if"
- " kalfilmsinputs_1stapp_ps->ztm1_track is -1. Please check this value.");
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
-
- return (0);
- }
-}
-//======================================================
-//= Integrating out the lagged regimes in order to
-//= updating zt_tm1 and Pt_tm1 for next perid tp1 through Kim-Nelson filter.
-//= tdata representing base-0 t timing, while inpt represents base-1 t timing.
-//
-//= Purpose: for each inpt, we integrate out grand regimes st
-//= only ONCE to prevent the dimension of updated zt_tm1 and Pt_tm1 through Kim-Nelson filter.
-//======================================================
-static int Updatekalfilms_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps)
-{
- //Output:
- // tm1update
- // z_{t_1+1|t_1}
- // P_{t_1+1|t_1}
- //Input:
- // t-1: base-1 t timing. Thus t-1=inpt-1.
-
- //--- Local variables
- int stp1i, sti, t_2, t_2p1;
- double prob_previous_regimes;
- //-- Output arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-(T+1).
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1).
- //--- Input arguments
- TSdcell *Gt_dc = kalfilmsinputs_1stapp_ps->Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //+
- TSdmatrix *bt_dm = kalfilmsinputs_1stapp_ps->bt_dm; //nz-by-nst.
- TSdcell *Ft_dc = kalfilmsinputs_1stapp_ps->Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Vt_dc = kalfilmsinputs_1stapp_ps->Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //+
- TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- int T = kalfilmsinputs_1stapp_ps->T;
- TSdvector z0_sdv;
- TSdvector btp1_sdv;
- TSdvector etdata_sdv;
- //=== Work arguments.
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- //+
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //=== For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(nz);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- //--- Critical checking.
- if (kalfilmsinputs_1stapp_ps->ztm1_track < 0)
- fn_DisplayError(".../kalman.c/Updatekalfilms_1stapp(): Make sure InitializeKalman_z10_P10() is called in the function RefreshEverthing()");
-
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- btp1_sdv.n = nz;
- btp1_sdv.flag = V_DEF;
- //+
- etdata_sdv.n = ny;
- etdata_sdv.flag = V_DEF;
-
- for (t_2=kalfilmsinputs_1stapp_ps->ztm1_track; t_2<t_1; t_2++)
- {
- //If t_1 <= ztm1_track, no updating.
- //If t_1 > ztm1_track, updating z_{t|t-1} and P_{t|t-1} up to t-1 = t_1.
-
- zt_tm1_dc->C[t_2p1=t_2+1]->flag = M_GE;
- for (stp1i=nst-1; stp1i>=0; stp1i--)
- {
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //=== Updating for next period by integrating out sti..
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i], PHtran_tdata_d4->F[t_2]->C[sti], 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_d4->F[t_2]->C[sti]);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- etdata_sdv.v = etdata_dc->C[t_2]->M + ny*sti;
- z0_sdv.v = zt_tm1_dc->C[t_2]->M + nz*sti; //sti: regime at time t_2.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, &etdata_sdv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i], Pt_tm1_d4->F[t_2]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
-
- //--- Integrating out the state at t_2 using
- //--- P(s_t_2|Y_{t_2}, theta) = ProbabilityStateConditionalCurrent(sti, t_2, smodel_ps);
- //--- One can also access to P(s_t_2|Y_{t_2}, theta) by using ElementV(smodel_ps->V[t_2],s_{t_2}i),
- //--- but this access will not call my function logTimetCondLH(), thus no updating for
- //--- P(s_t_2|Y_{t_2}, and thus leading to incorrect results.
- prob_previous_regimes = ProbabilityStateConditionalCurrent(sti, t_2, smodel_ps);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period.
- z0_sdv.v = zt_tm1_dc->C[t_2p1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[t_2p1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- //--- $$$ The following is important because it tells ProbabilityStateConditionalCurrent(), which calls
- //--- $$$ logTimetCondLH_kalfilms_1stapp(), which calls recursively this function again, that there is no
- //--- $$$ need to update Kalman filter for the period before kalfilmsinputs_1stapp_ps->ztm1_track.
- kalfilmsinputs_1stapp_ps->ztm1_track = t_2p1; //Means that z_{t_2p1+1|t_2p1} and P_{t_2p1+1|t_2p1} are done.
-
- //--- $$$ This function must be called after all the above computations are done.
- Update_et_Dt_1stapp(t_2p1, kalfilmsinputs_1stapp_ps);
- }
-
-
- //===
- DestroyMatrix_lf(Wnzbynz_dm);
- //
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (kalfilmsinputs_1stapp_ps->ztm1_track);
-}
-//======================================================
-//= Computes etdata and Dtdata for all grand regimes st at tbase0=inpt-1 or dtm1_track
-//= to prevent recomputing this object for different st at given tbase0.
-//======================================================
-static int Update_et_Dt_1stapp(int t_1, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps)
-{
- //Output:
- // dtm1_track is updated in this function.
- // PHtran_tdata_d4->F[t-1]
- // etdata_dc->C[t-1]
- // Dtdata_d4->F[t-1]
- //Input:
- // t_1=inpt-1: base-0 timing for et and Dt before the likelihood at time inpt is computed.
-
- //--- Local variables
- int sti, tbase0;
- //-- Output arguments
- TSdfourth *PHtran_tdata_d4 = kalfilmsinputs_1stapp_ps->PHtran_tdata_d4; //nz-by-ny-by-nst-T, saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc = kalfilmsinputs_1stapp_ps->etdata_dc; //ny-by-nst-by-T, save for computing the likelihood.
- TSdcell *yt_tm1_dc = kalfilmsinputs_1stapp_ps->yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0).
- TSdfourth *Dtdata_d4 = kalfilmsinputs_1stapp_ps->Dtdata_d4; //ny-by-ny-nst-by-T, save for computing the likelihood and updating Kalman filter Updatekalfilms_1stapp().
- //--- input arguments
- TSdcell *zt_tm1_dc = kalfilmsinputs_1stapp_ps->zt_tm1_dc; //nz-by-nst-by-T.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_1stapp_ps->Pt_tm1_d4; //nz-by-nz-by-nst-by-T.
- //+
- TSdmatrix *yt_dm = kalfilmsinputs_1stapp_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_1stapp_ps->at_dm; //ny-by-nst.
- TSdcell *Ht_dc = kalfilmsinputs_1stapp_ps->Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Rt_dc = kalfilmsinputs_1stapp_ps->Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- //--- Accessible variables
- int ny = kalfilmsinputs_1stapp_ps->ny;
- int nz = kalfilmsinputs_1stapp_ps->nz;
- int nst = kalfilmsinputs_1stapp_ps->nst;
- TSdvector z0_sdv;
- TSdvector yt_sdv, at_sdv;
- TSdvector etdata_sdv, yt_tm1_sdv;
- //=== Work arguments.
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- z0_sdv.n = nz;
- z0_sdv.flag = V_DEF;
- at_sdv.n = yt_sdv.n = ny;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- etdata_sdv.n = yt_tm1_sdv.n = ny;
- etdata_sdv.flag = yt_tm1_sdv.flag = V_DEF;
-
- for (tbase0=(kalfilmsinputs_1stapp_ps->dtm1_track+1); tbase0<=t_1; tbase0++)
- {
- //Note tbase0<=t_1, NOT tbase0<t_1.
- //If t_1 < (dtm1_track+1), no updating.
- //If t_1 >= (dtm1_track+1), updating etdata_dc->C[t-1] and Dtdata_d4->F[t-1] up to t-1=t_1.
-
- for (sti=nst-1; sti>=0; sti--)
- {
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti], 1.0, 0.0, 'N', 'T');
- CopyMatrix0(kalfilmsinputs_1stapp_ps->PHtran_tdata_d4->F[tbase0]->C[sti], PHtran_tdata_dm);
-
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti*at_dm->nrows; //grand regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- etdata_sdv.v = etdata_dc->C[tbase0]->M + etdata_sdv.n*sti;
- yt_tm1_sdv.v = etdata_dc->C[tbase0]->M + yt_tm1_sdv.n*sti;
- CopyVector0(&yt_tm1_sdv, &at_sdv);
- MatrixTimesVector(&yt_tm1_sdv, Ht_dc->C[sti], &z0_sdv, 1.0, 1.0, 'N'); //a + H*z_{t|t-1}.
- VectorMinusVector(&etdata_sdv, &yt_sdv, &yt_tm1_sdv); //y_t - a - H*z_{t|t-1}.
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
- CopyMatrix0(Dtdata_d4->F[tbase0]->C[sti], Dtdata_dm); //Saved to be used for logTimetCondLH_kalfilms_1stapp().
- }
-
- //--- $$$ This tracker functions the same way as kalfilmsinputs_1stapp_ps->ztm1_track.
- kalfilmsinputs_1stapp_ps->dtm1_track = tbase0;
- }
-
- //===
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (kalfilmsinputs_1stapp_ps->dtm1_track);
-}
-
-
-
-
-
-
-//-----------------------------------------------------
-//------------ OLD Code --------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//
-//- IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
-//- the coefficient regime is the 1st state variable, and
-//- the volatility regime is the 2nd state variable.
-//-----------------------------------------------------
-#if defined (NEWVERSIONofDW_SWITCH)
-double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0 (initial condition).
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int comst_c; //composite (s_tc, s_{t-1}c)
- int st_c, stm1_c, st_v;
- int comsti_c; //composite (s_tc, s_{t-1}c)
- int sti, sti_c, stm1i_c, sti_v;
- int comstp1i_c; //composite (s_{t+1}c, s_tc)
- int stp1i, stp1i_c, stp1i_v;
- int tbase0, tp1;
- double logdet_Dtdata, loglh_timet;
- static int record_tbase1_or_inpt_or_tp1 = 0;
- static int passonce;
- double prob_previous_regimes;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRstc = kalfilmsinputs_ps->nRstc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- int **Index = smodel_ps->sv->index; //Regime-switching states.
- //smodel_ps->sv->index is for our new code.
- // For old code (before 9 April 08 and before dsge_switch is created), use smodel_ps->sv->Index;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_logTimetLH_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- tbase0 = (tp1=inpt) - 1;
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- if (tbase0==0)
- {
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[comsti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[comsti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, comsti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[comsti_c], Ft_dc->C[comsti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
-//=== ???????? For debugging purpose.
-//if ((inpt<2) && (st==0))
-//{
-// fprintf(FPTR_DEBUG, "%%st=%d, inpt=%d, and sti=%d\n", st, inpt, sti);
-
-// fprintf(FPTR_DEBUG, "wP0_dv:\n");
-// WriteVector(FPTR_DEBUG, wP0_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Vt_dc->C[sti_v=%d]:\n", sti_v);
-// WriteMatrix(FPTR_DEBUG, Vt_dc->C[sti_v], " %10.5f ");
-
-// fflush(FPTR_DEBUG);
-
-//}
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime comsti_c=%d and sti_v=%d and at time=%d\n", comsti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
- }
-
-
- //======================================================
- //= Getting the logLH at time tbase0 or time inpt.
- //======================================================
- if (indxIndRegimes )
- {
- if (nRc==1) //Volatility.
- {
- comst_c = st_c = 0;
- st_v = st;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_v = Index[st][1]; //volatility state s_tv
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comst_c = st_c = st;
- st_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
- comst_c = st_c = Index[st][0];
- st_v = Index[st][1];
- }
- }
- else //Syncronized regimes
- {
- if (nRc>nRstc)
- {
- comst_c = Index[st][0]; //composite (s_tc, s_{t-1}c)
- st_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][0]; //coefficient regime at t.
- stm1_c = smodel_ps->sv->state_variable[0]->lag_index[comst_c][1]; //coefficient regime at t-1: tm1: t-1;
- st_v = st_c;
- }
- else
- comst_c = st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[comst_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comst_c*at_dm->nrows; //comst_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comst_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comst_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //--- Forming the log conditional likelihood at t.
- if (!isfinite(logdet_Dtdata=logdetspd(Dtdata_dm))) return (loglh_timet = -NEARINFINITY);
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
-//if ((inpt>82) && (inpt<86) )
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %.16e ");
-
-// fflush(FPTR_DEBUG);
-//}
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdet_Dtdata - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
-
-
-//=== ???????? For debugging purpose.
-if (inpt==1)
-{
- double wk1, wk2;
-
- wk1 = logdet_Dtdata;
- wk2 = VectorDotVector(wny_dv, etdata_dv);
- fprintf(FPTR_DEBUG, "logdet_Dtdata = %10.5f\n", wk1);
- fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f\n", wk2);
- fprintf(FPTR_DEBUG, "----- etdata_dv: \n");
- WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- yt_dv: \n");
- WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- at_dv: \n");
- WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- z0_dv: \n");
- WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
- fprintf(FPTR_DEBUG, "----- Ht_dc->C[comst_c=%d]:\n", comst_c);
- WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-
- fprintf(FPTR_DEBUG, "\n\n");
-
-}
-//
-fprintf(FPTR_DEBUG, " %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//fprintf(FPTR_DEBUG, "------------------------\n");
-//fprintf(FPTR_DEBUG, "st=%d and inpt=%d\n", st, inpt);
-//fprintf(FPTR_DEBUG, "loglh_timet = %10.5f\n", loglh_timet);
-//fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-//WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-////WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-////fprintf(FPTR_DEBUG, "\n");
-////WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//fflush(FPTR_DEBUG);
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>82) && (inpt<86) )
-//if (inpt<2)
-//{
-// //Must be declared at the top of this "if" block.
-// int kip1;
-// double tmp_Dtdata;
-// double tmp_expterm;
-
-// fprintf(FPTR_DEBUG, "%%------------------------\n");
-// fprintf(FPTR_DEBUG, "%%st=%d and inpt=%d\n", st, inpt);
-// fprintf(FPTR_DEBUG, "loglh_timet = %10.5f;\n", loglh_timet);
-
-
-// tmp_Dtdata = logdeterminant(Dtdata_dm);
-// tmp_expterm = VectorDotVector(wny_dv, etdata_dv);
-// fprintf(FPTR_DEBUG, "logdeterminant(Dtdata_dm) = %10.5f;\n", tmp_Dtdata);
-// fprintf(FPTR_DEBUG, "VectorDotVector(wny_dv, etdata_dv) = %10.5f;\n", tmp_expterm);
-// fprintf(FPTR_DEBUG, "wny_dv:\n");
-// WriteVector(FPTR_DEBUG, wny_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "etdata_dv:\n");
-// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&yt_sdv:\n");
-// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&at_sdv:\n");
-// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "&z0_sdv:\n");
-// WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-// fprintf(FPTR_DEBUG, "Ht_dc->C[comst_c=%d]:\n",comst_c);
-// WriteMatrix(FPTR_DEBUG, Ht_dc->C[comst_c], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Rt_dc->C[st_v=%d]:\n", st_v);
-// WriteMatrix(FPTR_DEBUG, Rt_dc->C[st_v], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4->F[tbase0]->C[st = %d]:\n",st);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[st], " %10.5f ");
-// fprintf(FPTR_DEBUG, "Dtdata_dm:\n");
-// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-
-
-
-
-//// WriteMatrix(FPTR_DEBUG, Dtdata_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "zt_tm1_dc->C[tbase0]:\n");
-//// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-//// //WriteVector(FPTR_DEBUG, &z0_sdv, " %10.5f ");
-//// //fprintf(FPTR_DEBUG, "\n");
-//// fprintf(FPTR_DEBUG, "bt_dm = [\n");
-//// WriteMatrix(FPTR_DEBUG, bt_dm, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-
-//// fprintf(FPTR_DEBUG, "et:\n");
-//// WriteVector(FPTR_DEBUG, etdata_dv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "yt_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &yt_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-//// fprintf(FPTR_DEBUG, "at_dv=[\n");
-//// WriteVector(FPTR_DEBUG, &at_sdv, " %10.5f ");
-//// fprintf(FPTR_DEBUG, "]';\n");
-
-
-//// for (ki=0; ki<Ht_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ht_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ht_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Ft_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Ft_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Ft_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-//// for (ki=0; ki<Vt_dc->ncells; ki++)
-//// {
-//// kip1 = ki+1;
-//// fprintf(FPTR_DEBUG, "Vt_dc(:,:,%d)=[\n", kip1);
-//// WriteMatrix(FPTR_DEBUG, Vt_dc->C[ki], " %10.5f ");
-//// fprintf(FPTR_DEBUG, "];\n");
-//// }
-// fflush(FPTR_DEBUG);
-//}
-
-
- //======================================================
- //= Updating zt_tm1 and Pt_tm1 for next perid tp1.
- //= tdata = tbase0 is base-0 timing.
- //======================================================
- if (inpt > record_tbase1_or_inpt_or_tp1) //This condition always satisfies at the 1st period (which is inpt=1).
- {
- passonce = 0;
- record_tbase1_or_inpt_or_tp1 = inpt;
- }
- if (!passonce)
- {
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comstp1i_c = stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_v = Index[stp1i][1]; //volatility state s_tv
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comstp1i_c = stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comstp1i_c = stp1i_c = Index[stp1i][0];
- stp1i_v = Index[stp1i][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comstp1i_c = Index[stp1i][0]; //composite (s_tc, s_{t-1}c)
- stp1i_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][0]; //coefficient regime at t.
- //sti_c = smodel_ps->sv->state_variable[0]->lag_index[comstp1i_c][1]; //coefficient regime at t-1: tm1: t-1;
- stp1i_v = stp1i_c;
- }
- else
- comstp1i_c = stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
-
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes)
- {
- if (nRc==1) //Volatility.
- {
- comsti_c = sti_c = 0;
- sti_v = sti;
- }
- else if ((nRv>1) && (nRc>nRstc)) //Trend inflation, both sc_t and sc_{t-1} enters coefficient regime.
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_v = Index[sti][1]; //volatility state s_tv
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- }
- else if ((nRv==1) && (nRc>nRstc))
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = 0;
- }
- else if ((nRv==1) && (nRc==nRstc))
- {
- comsti_c = sti_c = sti;
- sti_v = 0;
- }
- else if ((nRv>1) && (nRc==nRstc)) //only sc_t enters coefficient regime.
- {
- comsti_c = sti_c = Index[sti][0];
- sti_v = Index[sti][1];
- }
- }
- else //Syncronized regimes.
- {
- if (nRc>nRstc)
- {
- comsti_c = Index[sti][0]; //composite (s_tc, s_{t-1}c)
- sti_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][0]; //coefficient regime at t.
- stm1i_c = smodel_ps->sv->state_variable[0]->lag_index[comsti_c][1]; //coefficient regime at t-1: tm1: t-1;
- sti_v = sti_c;
- }
- else
- comsti_c = sti_c = sti_v = sti;
- }
-
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[comsti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + comsti_c*at_dm->nrows; //comsti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[comsti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[comsti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
- ScalarTimesMatrixSquare(Dtdata_dm, 0.5, Dtdata_dm, 'T', 0.5); //Making it symmetric against some rounding errors.
- //This making-symmetric is very IMPORTANT; otherwise, we will get the matrix being singular message
- // and eigenvalues being negative for the SPD matrix, etc. Then the likelihood becomes either
- // a bad number or a complex number.
- Dtdata_dm->flag = Dtdata_dm->flag | M_SU | M_SL;
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i_v]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_{t-1}, theta) = ElementV(smodel_ps->Z[inpt],s_{inpt}_i).
- //--- Note tbase0 = inpt-1 because the data in DW code (ElementV) is base-1.
- //--- Note at this point, we cannot access to P(s_t|Y_t, theta) = ElementV(smodel_ps->V[inpt],s_{inpt}_i)
- //--- through DW's code. But we can modify my own code to do this later.
- prob_previous_regimes = ElementV(smodel_ps->Z[inpt],sti);
- ScalarTimesVectorUpdate(ztp1_dv, prob_previous_regimes, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, prob_previous_regimes, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
- }
-
-
-//=== ???????? For debugging purpose.
-//if ((inpt>60) && (inpt<65) ) //if (inpt<5)
-//{
-// int kip1; //Must be declared at the top of this "if" block.
-
-// fprintf(FPTR_DEBUG, "zt_tm1t=[\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tbase0], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-
-// for (ki=0; ki<Pt_tm1_d4->F[tbase0]->ncells; ki++)
-// {
-// kip1 = ki+1;
-// fprintf(FPTR_DEBUG, "Pt_tm1_d4t(:,:,%d)=[\n", kip1);
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tbase0]->C[ki], " %10.5f ");
-// fprintf(FPTR_DEBUG, "];\n");
-// }
-
-// fflush(FPTR_DEBUG);
-//}
-
-
-//=== ???????? For debugging purpose.
-fprintf(FPTR_DEBUG, " loglh_timet = %10.5f\n", loglh_timet);
-fflush(FPTR_DEBUG);
-
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-#endif
-
-
-/**
-//----------------------------------------------------------------
-//-- Tested OK, but has not use because tz_Refresh_z_T7P_T_in_kalfilms_1st_approx()
-//-- cannot access to ElementV(smodel_ps->V[tp1],sti) or ElementV(smodel_ps->V[tbase0],sti)
-//-- because no likelihood has been formed at all before this function is called.
-//----------------------------------------------------------------
-#define LOG2PI (1.837877066409345e+000) //log(2*pi)
-//-----------------------------------------------------
-//- Updating or refreshing all Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// RunningGensys_const7varionly(lwzmodel_ps);
-// Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-//- before using tz_Refresh_z_T7P_T_in_kalfilms_1st_approx().
-//-----------------------------------------------------
-void tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- double debug1;
- //--- Local variables
- int stp1i, stp1i_c, stp1i_v, sti, sti_c, sti_v, tbase0, tp1;
- //=== Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int T = kalfilmsinputs_ps->T;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc = kalfilmsinputs_ps->Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm = kalfilmsinputs_ps->bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc = kalfilmsinputs_ps->Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc = kalfilmsinputs_ps->Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm = kalfilmsinputs_ps->z0_dm; //nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- TSdcell *P0_dc = kalfilmsinputs_ps->P0_dc; //nz-by-nz-by-nRc*nRv or nz-by-nRv, depending on indxIndRegimes.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- int nz2 = square(nz);
- TSdmatrix *Wnzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *Wnz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdmatrix *W2nz2bynz2_dm = CreateMatrix_lf(nz2,nz2);
- TSdvector *wP0_dv = CreateVector_lf(nz2);
- //+
- TSdvector yt_sdv, at_sdv, btp1_sdv; //zt_tm1_sdv, ztp1_t_sdv,
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *Wnzbyny_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *W2nzbynz_dm = CreateMatrix_lf(nz,nz);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
- TSdmatrix *Kt_tdata0_dm = CreateMatrix_lf(nz,ny);
- TSdmatrix *Kt_tdata_dm = CreateMatrix_lf(nz,ny);
- //--- For eigenvalue decompositions
- int ki;
- int errflag;
- double eigmax;
- TSdzvector *evals_dzv = evals_dzv = CreateVector_dz(nz);
- TSdvector *evals_abs_dv = CreateVector_lf(nz); //Absolute eigenvalues.
- //--- For updating zt_tm1_dm and Pt_tm1.
- TSdvector *ztp1_t_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_t_dm = CreateMatrix_lf(nz, nz);
- TSdvector *ztp1_dv = CreateVector_lf(z0_dm->nrows);
- TSdmatrix *Ptp1_dm = CreateMatrix_lf(nz, nz);
-
-
- if (smodel_ps->sv->nstates != z0_dm->ncols) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Make sure that the column dimension of z0_dm is the same as smodel_ps->sv->nstates");
- if (indxIndRegimes && (nRc>1) && (nRv>1))
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
-
-
- z0_sdv.n = z0_dm->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
- btp1_sdv.n = bt_dm->nrows;
- btp1_sdv.flag = V_DEF;
-
-
- //======= Initial condition. =======
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
-
- if (!kalfilmsinputs_ps->indxIni)
- {
- InitializeDiagonalMatrix_lf(Wnzbynz_dm, 1.0); //To be used for I(nz) -
- InitializeDiagonalMatrix_lf(Wnz2bynz2_dm, 1.0); //To be used for I(nz2) -
-
- //=== Eigenanalysis to determine the roots to ensure boundedness.
- errflag = eigrgen(evals_dzv, (TSdzmatrix *)NULL, (TSdzmatrix *)NULL, Ft_dc->C[sti_c]);
- if (errflag) fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): eigen decomposition failed");
- for (ki=nz-1; ki>=0; ki--) evals_abs_dv->v[ki] = sqrt(square(evals_dzv->real->v[ki]) + square(evals_dzv->imag->v[ki]));
- evals_abs_dv->flag = V_DEF;
- eigmax = MaxVector(evals_abs_dv);
- if (eigmax < (1.0+1.0e-14))
- {
- //--- Getting z0_dv: zt_tm1(:,1) = (eye(n_z)-F(:,:,1))\b(:,1);
- MatrixMinusMatrix(Wnzbynz_dm, Wnzbynz_dm, Ft_dc->C[sti_c]);
- z0_sdv.v = z0_dm->M + z0_sdv.n*sti;
- CopySubmatrix2vector(&z0_sdv, 0, bt_dm, 0, sti_c, bt_dm->nrows);
- bdivA_rgens(&z0_sdv, &z0_sdv, '\\', Wnzbynz_dm);
- //Done with Wnzbynz_dm.
- //--- Getting P0_dm: Pt_tm1(:,:,1) = reshape((eye(n_z^2)-kron(F(:,:,1),F(:,:,1)))\V1(:),n_z,n_z);
- tz_kron(W2nz2bynz2_dm, Ft_dc->C[sti_c], Ft_dc->C[sti_c]);
- MatrixMinusMatrix(Wnz2bynz2_dm, Wnz2bynz2_dm, W2nz2bynz2_dm);
- CopySubmatrix2vector(wP0_dv, 0, Vt_dc->C[sti_v], 0, 0, nz2);
- bdivA_rgens(wP0_dv, wP0_dv, '\\', Wnz2bynz2_dm);
- CopySubvector2matrix_unr(P0_dc->C[sti], 0, 0, wP0_dv, 0, nz2);
- //Done with all w*_dv and W*_dm.
- }
- else
- {
- fprintf(stdout, "\n-----------------\n");
- fprintf(stdout, "\nIn regime sti_c=%d and sti_v=%d and at time=%d\n", sti_c, sti_v, 0);
- fn_DisplayError("kalman.c/tz_Refresh_z_T7P_T_in_kalfilms_1st_approx(): the system is non-stationary solutions\n"
- " and the initial conditions must be supplied by, say, input arguments");
- fflush(stdout);
- }
- }
- }
- z0_dm->flag = M_GE;
- CopyMatrix0(zt_tm1_dc->C[0], z0_dm); //At time t=0.
- CopyCell0(Pt_tm1_d4->F[0], P0_dc); //At time t=0.
-
-
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[0]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[0]->C[0], " %.16e ");
-
-
- //============== Updating zt_tm1 and Pt_tm1. ==================
- for (tbase0=0; tbase0<T; tbase0++ )
- {
- //tdata = tbase0 is base-0 timing.
- tp1 = tbase0 + 1; //Next period.
-
- for (stp1i=smodel_ps->sv->nstates-1; stp1i>=0; stp1i--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- stp1i_c = 0;
- stp1i_v = stp1i;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- stp1i_c = stp1i;
- stp1i_v = 0;
- }
- else if (indxIndRegimes)
- {
- stp1i_c = smodel_ps->sv->Index[stp1i][0];
- stp1i_v = smodel_ps->sv->Index[stp1i][1];
- }
- else
- {
- stp1i_c = stp1i_v = stp1i;
- }
-
-
- InitializeConstantVector_lf(ztp1_dv, 0.0); //To be summed over sti.
- InitializeConstantMatrix_lf(Ptp1_dm, 0.0); //To be summed over sti.
- for (sti=smodel_ps->sv->nstates-1; sti>=0; sti--)
- {
- if (indxIndRegimes && (nRc==1))
- {
- sti_c = 0;
- sti_v = sti;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- sti_c = sti;
- sti_v = 0;
- }
- else if (indxIndRegimes)
- {
- sti_c = smodel_ps->sv->Index[sti][0];
- sti_v = smodel_ps->sv->Index[sti][1];
- }
- else
- {
- sti_c = sti_v = sti;
- }
-
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[sti], Ht_dc->C[sti_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + sti_c*at_dm->nrows; //sti_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*sti; //sti: regime at time tbase0.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[sti_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[sti_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[sti_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- //Done with z0_sdv.v.
-
-
- //=== Updating for next period by integrating out sti..
- if (tp1<T)
- {
- //Updating only up to tbase0=T-2. The values at tp1=T or tbase0=T-1 will not be used in the likelihood function.
-
- //--- Ktp1_t = (F_tp1*PHtran_t+G(:,:,t))/Dt;
- //--- Kt_tdata = (Ft*PHtran_tdata+G(:,:,tdata))/Dtdata where t=tp1 and tdata=t.
- CopyMatrix0(Kt_tdata0_dm, Gt_dc->C[sti_v]);
- MatrixTimesMatrix(Kt_tdata0_dm, Ft_dc->C[stp1i_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
- BdivA_rrect(Kt_tdata_dm, Kt_tdata0_dm, '/', Dtdata_dm);
- //+ zt_tm1(:,t) = b(:,t) + Ft*zt_tm1(:,tdata) + Kt_tdata*etdata where t=tp1 and tm1=t.
- MatrixTimesVector(ztp1_t_dv, Ft_dc->C[stp1i_c], &z0_sdv, 1.0, 0.0, 'N');
- MatrixTimesVector(ztp1_t_dv, Kt_tdata_dm, etdata_dv, 1.0, 1.0, 'N');
- btp1_sdv.v = bt_dm->M + stp1i_c*btp1_sdv.n;
- VectorPlusMinusVectorUpdate(ztp1_t_dv, &btp1_sdv, 1.0);
- //+ Pt_tm1(:,:,t) = Ft*Ptdata*Fttran - Kt_tdata*Dtdata*Kt_tdatatran + V(:,:,t);
- CopyMatrix0(Ptp1_t_dm, Vt_dc->C[stp1i]);
- MatrixTimesMatrix(Wnzbyny_dm, Kt_tdata_dm, Dtdata_dm, 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(Wnzbynz_dm, Wnzbyny_dm, Kt_tdata_dm, 1.0, 0.0, 'N', 'T');
- MatrixPlusMinusMatrixUpdate(Ptp1_t_dm, Wnzbynz_dm, -1.0);
- //Done with all W*_dm.
- MatrixTimesMatrix(Wnzbynz_dm, Ft_dc->C[stp1i_c], Pt_tm1_d4->F[tbase0]->C[sti], 1.0, 0.0, 'N', 'N');
- MatrixTimesMatrix(W2nzbynz_dm, Wnzbynz_dm, Ft_dc->C[stp1i_c], 1.0, 0.0, 'N', 'T');
- MatrixPlusMatrixUpdate(Ptp1_t_dm, W2nzbynz_dm);
- //Done with all W*_dm.
-
- //--- Integrating out the state at tbase0 using P(s_t|Y_t, theta) = ElementV(smodel_ps->V[t+1],s_{t+1}_i).
- //--- Note because the data in DW code (ElementV) is base-1, t+1 is actually tbase0.
- debug1 = ElementV(smodel_ps->V[tp1],sti); //?????? Debug.
- //ScalarTimesVectorUpdate(ztp1_dv, ElementV(smodel_ps->V[tp1],sti), ztp1_t_dv);
- //ScalarTimesMatrix(Ptp1_dm, ElementV(smodel_ps->V[tp1],sti), Ptp1_t_dm, 1.0);
- ScalarTimesVectorUpdate(ztp1_dv, 0.5, ztp1_t_dv);
- ScalarTimesMatrix(Ptp1_dm, 0.5, Ptp1_t_dm, 1.0);
- Ptp1_dm->flag = M_GE | M_SU | M_SL;
- //Done with ztp1_t_dv and Ptp1_t_dm.
- }
- }
- //--- Filling zt_tm1 and Pt_tm1 for next period
- if (tp1<T)
- {
- z0_sdv.v = zt_tm1_dc->C[tp1]->M + z0_sdv.n*stp1i; //stp1i: regime at time tp1.
- CopyVector0(&z0_sdv, ztp1_dv);
- CopyMatrix0(Pt_tm1_d4->F[tp1]->C[stp1i], Ptp1_dm); //stp1i: regime at time tp1.
- //Done with ztp1_dv, z0_sdv, Ptp1_dm.
- }
- }
- if (tp1<T)
- zt_tm1_dc->C[tp1]->flag = M_GE;
-
-// fprintf(FPTR_DEBUG, "\n &yt_sdv:\n");
-// WriteMatrix(FPTR_DEBUG, &yt_sdv, " %.16e ");
-// fprintf(FPTR_DEBUG, "\n zt_tm1_dc->C[tp1]:\n");
-// WriteMatrix(FPTR_DEBUG, zt_tm1_dc->C[tp1], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fprintf(FPTR_DEBUG, "\n Pt_tm1_d4->F[tp1]->C[0]:\n");
-// WriteMatrix(FPTR_DEBUG, Pt_tm1_d4->F[tp1]->C[0], " %.16e ");
-// fprintf(FPTR_DEBUG, "\n");
-// fflush(FPTR_DEBUG);
-
-
- }
-
- //===
- DestroyVector_dz(evals_dzv);
- DestroyVector_lf(evals_abs_dv);
- DestroyMatrix_lf(Wnzbynz_dm);
- DestroyMatrix_lf(Wnz2bynz2_dm);
- DestroyMatrix_lf(W2nz2bynz2_dm);
- DestroyVector_lf(wP0_dv);
- //
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(Wnzbyny_dm);
- DestroyMatrix_lf(W2nzbynz_dm);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
- DestroyMatrix_lf(Kt_tdata0_dm);
- DestroyMatrix_lf(Kt_tdata_dm);
- //
- DestroyVector_lf(ztp1_t_dv);
- DestroyMatrix_lf(Ptp1_t_dm);
- DestroyVector_lf(ztp1_dv);
- DestroyMatrix_lf(Ptp1_dm);
-}
-//-----------------------------------------------------
-//- Kalman filter at time t for Markov-switching DSGE model.
-//- WARNING: make sure to call the following functions
-// (1) RunningGensys_const7varionly(lwzmodel_ps);
-// (2) Refresh_kalfilms_*(lwzmodel_ps); //Creates or refreshes kalfilmsinputs_ps at new parameter values.
-// (3) tz_Refresh_z_T7P_T_in_kalfilms_1st_approx();
-//- before using kalfilms_timet_1st_approx().
-//-----------------------------------------------------
-double tz_kalfilms_timet_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps)
-{
- //st, st_c, and st_v: base-0: deals with the cross-section values at time t where
- // st is a grand regime, st_c is an encoded coefficient regime, and st_c is an encoded volatility regime.
- //inpt: base-1 in the sense that inpt>=1 to deal with the time series situation where S_T is (T+1)-by-1 and Y_T is T+nlags_max-by-1.
- // The 1st element for S_T is S_T[1] while S_T[0] is s_0. The same for (T+1)-by-1 gbeta_dv and nlcoefs-by-(T+1) galpha_dm.
- // The 1st element for Y_T, however, is Y_T[nlags_max+1-1].
- //See (42.3) on p.42 in the SWZII NOTES.
-
-
- //--- Local variables
- int st_c, st_v, tbase0;
- double loglh_timet;
- //--- Accessible variables
- int ny = kalfilmsinputs_ps->ny;
- int nz = kalfilmsinputs_ps->nz;
- int nRc = kalfilmsinputs_ps->nRc;
- int nRv = kalfilmsinputs_ps->nRv;
- int indxIndRegimes = kalfilmsinputs_ps->indxIndRegimes;
- TSdvector z0_sdv;
- //+ input arguments.
- TSdmatrix *yt_dm = kalfilmsinputs_ps->yt_dm; //ny-by-T.
- TSdmatrix *at_dm = kalfilmsinputs_ps->at_dm; //ny-by-nRc.
- TSdcell *Ht_dc = kalfilmsinputs_ps->Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc = kalfilmsinputs_ps->Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- //+ Output arguments.
- TSdcell *zt_tm1_dc = kalfilmsinputs_ps->zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4 = kalfilmsinputs_ps->Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- //=== Work arguments.
- TSdvector yt_sdv, at_sdv;
- TSdvector *wny_dv = CreateVector_lf(ny);
- TSdmatrix *PHtran_tdata_dm = CreateMatrix_lf(nz,ny);
- TSdvector *etdata_dv = CreateVector_lf(ny);
- TSdmatrix *Dtdata_dm = CreateMatrix_lf(ny,ny);
-
-
- if (smodel_ps->sv->nstates != zt_tm1_dc->C[0]->ncols) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Make sure that the column dimension of zt_tm1_dc->C is the same as smodel_ps->sv->nstates");
-
- tbase0 = inpt - 1; //base-0 time t.
-
- if (indxIndRegimes && (nRc==1))
- {
- st_c = 0;
- st_v = st;
- }
- else if (indxIndRegimes && (nRv==1))
- {
- st_c = st;
- st_v = 0;
- }
- else if (indxIndRegimes)
- {
- if (smodel_ps->sv->n_state_variables != 2) fn_DisplayError("kalman.c/kalfilms_timet_1st_approx():\n"
- " Number of state variables must be coincide with indxIndRegimes");
- st_c = smodel_ps->sv->Index[st][0];
- st_v = smodel_ps->sv->Index[st][1];
- }
- else
- {
- st_c = st_v = st;
- }
-
-
- z0_sdv.n = zt_tm1_dc->C[0]->nrows;
- z0_sdv.flag = V_DEF;
- //
- at_sdv.n = yt_sdv.n = yt_dm->nrows;
- at_sdv.flag = yt_sdv.flag = V_DEF;
-
- //====== Computing the conditional LH at time t. ======
- //--- Setup.
- MatrixTimesMatrix(PHtran_tdata_dm, Pt_tm1_d4->F[tbase0]->C[st], Ht_dc->C[st_c], 1.0, 0.0, 'N', 'T');
-
- //--- Data.
- //- etdata = Y_T(:,tdata) - a(:,tdata) - Htdata*ztdata where tdata = tbase0 = inpt-1.
- yt_sdv.v = yt_dm->M + tbase0*yt_dm->nrows;
- at_sdv.v = at_dm->M + st_c*at_dm->nrows; //st_c: coefficient regime at time tbase0.
- z0_sdv.v = zt_tm1_dc->C[tbase0]->M + z0_sdv.n*st; //st: regime at time tbase0 for zt_tm1.
- VectorMinusVector(etdata_dv, &yt_sdv, &at_sdv);
- MatrixTimesVector(etdata_dv, Ht_dc->C[st_c], &z0_sdv, -1.0, 1.0, 'N');
- //+ Dtdata = Htdata*PHtran_tdata + R(:,:,tbase0);
- CopyMatrix0(Dtdata_dm, Rt_dc->C[st_v]);
- MatrixTimesMatrix(Dtdata_dm, Ht_dc->C[st_c], PHtran_tdata_dm, 1.0, 1.0, 'N', 'N');
-
- //--- Forming the log conditional likelihood at t.
- bdivA_rgens(wny_dv, etdata_dv, '/', Dtdata_dm);
- loglh_timet = -(0.5*ny)*LOG2PI - 0.5*logdeterminant(Dtdata_dm) - 0.5*VectorDotVector(wny_dv, etdata_dv);
- //Done with all w*_dv.
-
-
- //===
- DestroyVector_lf(wny_dv);
- DestroyMatrix_lf(PHtran_tdata_dm);
- DestroyVector_lf(etdata_dv);
- DestroyMatrix_lf(Dtdata_dm);
-
- return (loglh_timet);
-}
-#undef LOG2PI
-/**/
-
-
-
-
diff --git a/CFiles/kalmanOldWorks2.h b/CFiles/kalmanOldWorks2.h
deleted file mode 100755
index df661b5160ecd1be8c733931ddfc94fb29d2f7dd..0000000000000000000000000000000000000000
--- a/CFiles/kalmanOldWorks2.h
+++ /dev/null
@@ -1,314 +0,0 @@
-#ifndef __KALMAN_H__
- #define __KALMAN_H__
-
- #include "tzmatlab.h"
- #include "mathlib.h"
- #include "switch.h"
- #include "fn_filesetup.h" //Used to call WriteMatrix(FPTR_DEBUG,....).
-
-
- typedef struct TSkalcvfurw_tag {
- //urw: univariate random walk kalman filter. Desigend specially for the 2006 AER SWZ paper.
-
- //=== Input arguments.
- int indx_tvsigmasq; //0: constant siqmasq in Kalman updating (default);
- //1: Keyensian (project-specific) type of time-varying sigmasq in Kalman updating; See pp.37 and 37a in SWZ Learning NOTES;
- //2: project-specific type;
- //3: another project-specific type.
- double sigmasq; //Variance for the residual eps(t) of the measurement equation.
- int fss; //T: effective sample size (excluding lags).
- int kx; //dimension for x(t).
- TSdmatrix *V_dm; //kx-by-kx. Covariance (symmetric and positive definite) matrix for the residual eta(t) of the transition equation.
- TSdvector *ylhtran_dv; //1-by-T of y(t). The term lh means lelf hand side and tran means transpose.
- TSdmatrix *Xrhtran_dm; //kx-by-T of x(t). The term rh means right hand side and tran means transpose.
- TSdvector *z10_dv; //kx-by-1. Initial condition for prediction: z_{1|0}.
- TSdmatrix *P10_dm; //kx-by-kx symmetric matrix. Initial condition for the variance of the prediction: P_{1|0}.
-
- //=== Output arguments.
- TSdvector *zupdate_dv; //kx-by-1. z_{T+1|T}.
- TSdmatrix *Zpredtran_dm; //kx-by-T matrix of one-step predicted values of z(t). [z_{2|1}, ..., z_{t+1|t}, ..., z_{T+1|T}].
- //Set to NULL (no output) if storeZ = 0;
- TSdcell *Ppred_dc; //T cells and kx-by-kx symmetric and positive definite matrix for each cell. Mean square errors of predicted state variables.
- //{P_{2|1}, ..., P{t+1|t}, ..., P{T+1|T}. Set to NULL (no output if storeV = 0).
- TSdvector *ylhtranpred_dv; //1-by-T one-step prediction of y(t) or ylhtran_dv. Added 03/17/05.
-
- //=== Function itself.
- void (*learning_fnc)(struct TSkalcvfurw_tag *, void *);
- } TSkalcvfurw; //urw: univariate random walk.
- //
- typedef void TFlearninguni(struct TSkalcvfurw_tag *, void *); //For linear rational expectations models.
-
-
- //=== Better version is TSkalfilmsinputs_1stapp_tag. Kalman filter for constant or known-time-varying DSGE models.
- typedef struct TSkalfiltv_tag
- {
- //General (known-time-varying) Kalman filter for DSGE models.
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs are as follows:
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-T matrix of time-varying input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-T 3-D of time-varying matrices in the measurement equation.
- // R is an n_y-by-n_y-by-T 3-D of time-varying covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-T 3-D of time-varying E(eta_t * eps_t').
- // ------
- // b is an n_z-by-T matrix of time-varying input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-T 3-D of time-varying transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-T 3-D of time-varying covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-1 vector of initial condition when indxIni=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z matrix of initial condition when indxIni=1. (Not used if indxIni=0.)
- //
- // Outputs are as follows:
- // loglh is a value of the log likelihood function of the state-space model
- // under the assumption that errors are multivariate Gaussian.
- // zt_tm1 is an n_z-by-T matrices of one-step predicted state vectors with z0_0m1 as a initial condition
- // and with z_{t+1|t} as the last element. Thus, we can use it as a base-1 vector.
- // Pt_tm1 is an n_z-by-n_z-by-T 3-D of covariance matrices of zt_tm1 with P0_0m1 as a initial condition
- // and with P_{t+1|t} as the last element. Thus, we can use it as a base-1 cell.
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // March 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-T.
- TSdcell *Ht_dc; //ny-by-nz-by-T.
- TSdcell *Rt_dc; //ny-by-ny-by-T. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-T. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-T.
- TSdcell *Ft_dc; //nz-by-nz-by-T.
- TSdcell *Vt_dc; //nz-by-nz-by-T. Covariance matrix for the state equation.
- //
- TSdvector *z0_dv; //nz-by-1;
- TSdmatrix *P0_dm; //nz-by-nz.
-
- //=== Output arguments.
- double loglh; //log likelihood.
- TSdmatrix *zt_tm1_dm; //nz-by-T.
- TSdcell *Pt_tm1_dc; //nz-by-nz-T.
- } TSkalfiltv;
-
-
-
- //=== Inputs for filter for Markov-switching DSGE models at any time t.
- typedef struct TSkalfilmsinputs_1stapp_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on the grand regime s_t that follows a Markov-chain process
- // and is taken as given, and
- // eps(t) = Psi_u(t)*u(t), where Psi_u(t) is n_y-by-n_u and u(t) is n_u-by-1;
- // eta(t) = Psi_e(t)*e(t), where Psi_e(t) is n_z-by-n_e and e(t) is n_e-by-1.
- //
- // Inputs at time t are as follows where nst is number of grand regimes (including lagged regime
- // and coefficients and shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nst matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nst 3-D of Markov-switching matrices in the measurement equation.
- // Psi_u is an n_y-by-n_u-by-nst 3-D Markov-switching matrices.
- // R is an n_y-by-n_y-by-nst 3-D of Markov-switching covariance matrices,
- // E(eps(t) * eps(t)') = Psi_u(t)*Psi_u(t)', for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nst 3-D of Markov-switching E(eta_t * eps_t') = Psi_e(t)*Psi_u(t)'.
- // ------
- // b is an n_z-by-nst matrix of Markov-switching input vectors in the state equation with b(:,st) as an initial condition.
- // (alternatively, with the ergodic weighted b(:,st) as an initial condition).
- // F is an n_z-by-n_z-by-nst 3-D of Markov-switching transition matrices in the state equation with F(:,:,st)
- // as an initial condition (alternatively, with the ergodic weighted F(:,:,st) as an initial condition).
- // Psi_e is an n_z-by-n_e-by-nst 3-D Markov-switching matrices.
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices,
- // E(eta(t)*eta(t)') = Psi_e(t)*Psi_e(t)', for the error in the state equation
- // with V(:,:,st) as an initial condition (alternatively, with the ergodic weighted V(:,:,st) as an initial condition).
- // ------
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nst matrix of initial condition (Not used if indxIni=0).
- // P0 is an n_z-by-n_z-by-nst 3-D of initial condition (Not used if indxIni=0).
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,st))\b(:,st)
- // vec(P0_0m1) = (I-kron(F(:,:,st),F(:,:,st)))\vec(V(:,:,st))
- // Note that all eigenvalues of the matrix F(:,:,st) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha. Revised, April 2008.
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nu; //number of measurement errors.
- int ne; //number of fundamental errors.
- int nst; //number of grand composite regimes (current and past regimes, coefficient and volatility regimes).
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0,
- //0: using the unconditional momnets for any given regime at time 0 (default when indxDiffuse = 0).
- int indxDiffuse; //1: using the diffuse condition for z_{1|0} and P_{1|0} (default option), according to Koopman and Durbin, "Filtering and Smoothing of State Vector for Diffuse State-Space Models," J. of Time Series Analysis, Vol 24(1), pp.85-99.
- //0: using the unconditional moments.
- double DiffuseScale; //A large (infinity) number when indxDiffuse = 1.
- int ztm1_track; //t-1 = -1: no initial conditions z_{1|0} and P_{1|0} has been computed yet, but will be using InitializeKalman_z10_P10(),
- //t-1 >= 0:T-1: z_{t|t-1} and P_{t|t-1} are updated up to t-1.
- int dtm1_track; //t-1 = -1: no etdata_dc->C[0] or Dtdata_d4->F[0] has been computed yet.
- //t-1 >= 0:T-1: etdata_dc->C[t-1] and Dtdata_d4->F[t-1] are updated up to t-1.
-
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nst.
- TSdcell *Ht_dc; //ny-by-nz-by-nst.
- TSdcell *Psiut_dc; //ny-by-nu-by-nst. Measurement error coefficient matrix.
- TSdcell *Rt_dc; //ny-by-ny-by-nst. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nst. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nst.
- TSdcell *Ft_dc; //nz-by-nz-by-nst.
- TSdcell *Psiet_dc; //nz-by-ne-by-nst. Impact matrix in the state equation.
- TSdcell *Vt_dc; //nz-by-nz-by-nst. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_0_dm; //nz-by-nst. z_{0|0}.
- TSdmatrix *z0_dm; //nz-by-nst. z_{1|0}.
- TSdcell *P0_dc; //nz-by-nz-by-nst. P_{1|0}
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nst-by-(T+1), where z_{1|0} is an initial condition (1st element with t-1=0 or t=1 for base-1) and
- // the terminal condition z_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, z_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nst-by-(T+1), where P_{1|0} is an initial condition (1st element with t-1=0) and
- // the terminal condition P_{T+1|T} using Updatekalfilms_1stapp(T, ...) is not computed
- // when the likelihood logTimetCondLH_kalfilms_1stapp() is computed. Thus, P_{T+1|T}
- // has not legal value computed in most applications unless in out-of-sample forecasting problems.
- //+ Will be save for updating likelihood and Kalman filter Updatekalfilms_1stapp(), so save time to recompute these objects again.
- TSdfourth *PHtran_tdata_d4; //nz-by-ny-by-nst-T, P_{t|t-1}*H_t'. Saved only for updating Kalman filter Updatekalfilms_1stapp().
- TSdcell *etdata_dc; //ny-by-nst-by-T (with base-0 T), forecast errors e_t in the likelihood.
- TSdcell *yt_tm1_dc; //ny-by-nst-by-T, one-step forecast y_{t|t-1} for t=0 to T-1 (base-0). Incorrect now (Used to back out structural shocks).
- TSdfourth *Dtdata_d4; //ny-by-ny-nst-by-T, forecast covariance D_t in the likelihood. Saved for updating Kalman filter Updatekalfilms_1stapp().
- } TSkalfilmsinputs_1stapp;
-
-
- //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- //~~~ OLD Code: Inputs for filter for Markov-switching DSGE models at any time t.
- //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- typedef struct TSkalfilmsinputs_tag
- {
- //Inputs Markov-switching Kalman filter for DSGE models (conditional on all the regimes at time t).
- // It computes a sequence of one-step predictions and their covariance matrices, and the log likelihood.
- // The function uses a forward recursion algorithm. See also the Matlab function fn_kalfil_tv.m
- //
- // State space model is defined as follows:
- // y(t) = a(t) + H(t)*z(t) + eps(t) (observation or measurement equation)
- // z(t) = b(t) + F(t)*z(t) + eta(t) (state or transition equation)
- // where a(t), H(t), b(t), and F(t) depend on s_t that follows a Markov-chain process and are taken as given.
- //
- // Inputs at time t are as follows where nRc is number of regimes for coefficients
- // nRv is number of regimes for volatility (shock variances):
- // Y_T is a n_y-by-T matrix containing data [y(1), ... , y(T)].
- // a is an n_y-by-nRc matrix of Markov-switching input vectors in the measurement equation.
- // H is an n_y-by-n_z-by-nRc 3-D of Markov-switching matrices in the measurement equation.
- // R is an n_y-by-n_y-by-nRv 3-D of Markov-switching covariance matrices for the error in the measurement equation.
- // G is an n_z-by-n_y-by-nRv 3-D of Markov-switching E(eta_t * eps_t').
- // ------
- // b is an n_z-by-nRc matrix of Markov-switching input vectors in the state equation with b(:,1) as an initial condition.
- // F is an n_z-by-n_z-by-nRc 3-D of Markov-switching transition matrices in the state equation with F(:,:,1) as an initial condition.
- // V is an n_z-by-n_z-by-nRv 3-D of Markov-switching covariance matrices for the error in the state equation with V(:,:,1) as an initial condition.
- // ------
- // indxIndRegimes: 1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- // indxIni: 1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- // 0: using the unconditional mean for any given regime at time 0.
- // z0 is an n_z-by-nRc*nRv matrix of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // z0 is an n_z-by-nRv matrix of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRc*nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=1. (Not used if indxIni=0.)
- // P0 is an n_z-by-n_z-by-nRv 3-D of initial condition when indxIni=1 and indxIndRegimes=0. (Not used if indxIni=0.)
- //
- // The initial state vector and its covariance matrix are computed under the bounded (stationary) condition:
- // z0_0m1 = (I-F(:,:,1))\b(:,1)
- // vec(P0_0m1) = (I-kron(F(:,:,1),F(:,:,1)))\vec(V(:,:,1))
- // Note that all eigenvalues of the matrix F(:,:,1) are inside the unit circle when the state-space model is bounded (stationary).
- //
- // November 2007, written by Tao Zha
- // See Hamilton's book ([13.2.13] -- [13.2.22]), Harvey (pp.100-106), and LiuWZ Model I NOTES pp.001-003.
-
- //=== Input arguments.
- int ny; //number of observables.
- int nz; //number of state variables.
- int nRc; //number of composite regimes (current and past regimes) for coefficients.
- int nRstc; //number of coefficient regimes at time t.
- int nRv; //number of regimes for volatility (shock variances).
- int indxIndRegimes; //1: coefficient regime and volatility regime are independent; 0: these two regimes are synchronized completely.
- int T; //sample size.
- int indxIni; //1: using the initial condition with zt_tm1(:,1)=z0 and Pt_tm1(:,:,1)=P0;
- //0: using the unconditional mean for any given regime at time 0. (Default value)
- TSdmatrix *yt_dm; //ny-by-T.
- TSdmatrix *at_dm; //ny-by-nRc.
- TSdcell *Ht_dc; //ny-by-nz-by-nRc.
- TSdcell *Rt_dc; //ny-by-ny-by-nRv. Covariance matrix for the measurement equation.
- TSdcell *Gt_dc; //nz-by-ny-by-nRv. Cross-covariance.
- //
- TSdmatrix *bt_dm; //nz-by-nRc.
- TSdcell *Ft_dc; //nz-by-nz-by-nRc.
- TSdcell *Vt_dc; //nz-by-nz-by-nRv. Covariance matrix for the state equation.
- //
- TSdmatrix *z0_dm; //nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nRv if indxIndRegimes == 0.
- TSdcell *P0_dc; //nz-by-nz-by-nRc*nRv if indxIndRegimes == 1 or nz-by-nz-by-nRv if indxIndRegimes == 0.
-
-
- //=== Output arguments only used for 1st order approximation to zt and Pt depending on infinite past regimes.
- TSdcell *zt_tm1_dc; //nz-by-nRc*nRv-by-T if indxIndRegimes==1, nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- TSdfourth *Pt_tm1_d4; //nz-by-nz-by-nRc*nRv-T if indxIndRegimes==1, nz-by-nz-by-nRv-by-T if indxIndRegimes==0 where nRc=nRv.
- } TSkalfilmsinputs;
-
-
-
-
- //--- Functions for univariate random walk kalman filter.
- TSkalcvfurw *CreateTSkalcvfurw(TFlearninguni *func, int T, int k, int tv); //, int storeZ, int storeV);
- TSkalcvfurw *DestroyTSkalcvfurw(TSkalcvfurw *kalcvfurw_ps);
- void kalcvf_urw(TSkalcvfurw *kalcvfurw_ps, void *dummy_ps);
-
- //--- New Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp2(int ny, int nz, int nu, int ne, int nst, int T);
- struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp2(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
- struct TSkalfilmsinputs_1stapp_tag *CreateTSkalfilmsinputs_1stapp(int ny, int nz, int nst, int T);
- struct TSkalfilmsinputs_1stapp_tag *DestroyTSkalfilmsinputs_1stapp(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps);
- int InitializeKalman_z10_P10(struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, TSdmatrix *z10_dm, TSdcell *P10_dc);
- double logTimetCondLH_kalfilms_1stapp(int st, int inpt, struct TSkalfilmsinputs_1stapp_tag *kalfilmsinputs_1stapp_ps, struct TStateModel_tag *smodel_ps);
-
-
-
- //--- OLD Code: Functions for general constant Kalman filter.
- struct TSkalfiltv_tag *CreateTSkalfiltv(int ny, int nz, int T);
- struct TSkalfiltv_tag *DestroyTSkalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- //Used to test tz_logTimetCondLH_kalfiltv(). (Done April 08). double tz_kalfiltv(struct TSkalfiltv_tag *kalfiltv_ps);
- double tz_logTimetCondLH_kalfiltv(int st, int inpt, struct TSkalfiltv_tag *kalfiltv_ps);
-
- //--- OLD Code: Functions for Markov-switching Kalman filter.
- struct TSkalfilmsinputs_tag *CreateTSkalfilmsinputs(int ny, int nz, int nRc, int nRstc, int nRv, int indxIndRegimes, int T);
- struct TSkalfilmsinputs_tag *DestroyTSkalfilmsinputs(struct TSkalfilmsinputs_tag *kalfilmsinputs_ps);
- double tz_logTimetCondLH_kalfilms_1st_approx(int st, int inpt, struct TSkalfilmsinputs_tag *kalfilmsinputs_ps, struct TStateModel_tag *smodel_ps);
- //IMPORTANT NOTE: in the Markov-switching input file datainp_markov*.prn, it MUST be that
- // the coefficient regime is the 1st state variable, and
- // the volatility regime is the 2nd state variable.
-#endif
-
diff --git a/CFiles/mathlib.c b/CFiles/mathlib.c
index 8bf1c40cdd0ae6052d2e6bb73070b00171462171..09a044992097e86cc47b8801116ecc93999b5dd6 100755
--- a/CFiles/mathlib.c
+++ b/CFiles/mathlib.c
@@ -3835,6 +3835,49 @@ void CopySubmatrix0(TSdmatrix *x1_dm, TSdmatrix *x2_dm, const int br, const int
else fn_DisplayError(".../mathlib.c/CopySubmatrix0(): the submatrix of x2_dm must be within the ranges of both x1_dm and x2_dm");
}
+void CopySubmatrixSelectedCols0(TSdmatrix *x1_dm, TSdmatrix *x2_dm, const int br, const int nrs, TSivector *locs4cols)
+{
+ //Copies the nrs-by-locs4cols->n submatrix of x2_dm to the most left corner of x1_dm (i.e., at 0).
+ //Note: br means the beginning of the row (*must* be 0 based) for this submatrix of x2_dm, inclusive;
+ // locs4cols indicates the locations of columns (*must* be 0 based) for this submatrix of x2_dm to be pasted, inclusive.
+ int _j, loc1, loc2, ncols_copy,
+ nrows1, ncols1, nrows2, ncols2;
+
+ if ( !x1_dm || !x2_dm || !locs4cols) fn_DisplayError(".../mathlib.c/CopySubmatrixSelectedCols0(): All input matrices must be created (memory-allocated)");
+ else if ( !x2_dm->flag || !locs4cols->flag ) fn_DisplayError(".../mathlib.c/CopySubmatrixSelectedCols0(): R input matrix must be given values");
+ else {
+ nrows1=x1_dm->nrows;
+ ncols1=x1_dm->ncols;
+ nrows2=x2_dm->nrows;
+ ncols2=x2_dm->ncols;
+ }
+ if (MaxVector_int(locs4cols)>ncols2)
+ fn_DisplayError(".../mathlib.c/CopySubmatrixSelectedCols0(): selected columns in x2_dm cannot exceed the number of columns of x2_dm");
+
+ if ( !(x2_dm->flag & M_GE) ) {
+ if (x2_dm->flag & M_SU) SUtoGE(x2_dm);
+ else if (x2_dm->flag & M_SL) SLtoGE(x2_dm);
+ else fn_DisplayError(".../mathlib.c/CopySubmatrixSelectedCols0(): Haven't got time to deal with the M_UT and M_LT cases for x2_dm");
+ }
+
+ //=== Performs the operation.
+ if ( ((br+nrs)<=nrows2) && ((ncols_copy=locs4cols->n)<=ncols1) && (nrs<=nrows1) )
+ {
+ for (_j=ncols_copy-1; _j>=0; _j--) {
+ loc1 = _j*nrows1; //Points to the top of the column in x1_dm.
+ loc2 = mos(br, locs4cols->v[_j], nrows2); //Points to the top of the column in the submatrix of x2_dm.
+ #if defined(SWITCHTOINTELCMATH) // define: use Intek MKL LAPACK library; undef: use others.
+ cblas_dcopy(nrs, x2_dm->M+loc2, 1, x1_dm->M+loc1, 1);
+ #elif defined(SWITCHTOTZCMATH) // define: use my own C math library ; undef: use others.
+ memcpy(x1_dm->M+loc1, x2_dm->M+loc2, nrs*sizeof(double));
+ #endif
+
+ }
+ x1_dm->flag = M_GE;
+ }
+ else fn_DisplayError(".../mathlib.c/CopySubmatrixSelectedCols0(): the submatrix of x2_dm must be within the ranges of both x1_dm and x2_dm");
+}
+
void CopySubmatrix(TSdmatrix *x1_dm, const int br1, const int bc1, TSdmatrix *x2_dm, const int br2, const int bc2, const int nrs, const int ncs) {
//Copies the nrs-by-ncs submatrix of x2_dm to x1_dm at the specified location (br1, bc1).
//Note: br1 means the beginning of the row (*must* be 0 based) for x1_dm, inclusive.
@@ -3864,6 +3907,7 @@ void CopySubmatrix(TSdmatrix *x1_dm, const int br1, const int bc1, TSdmatrix *x2
else fn_DisplayError(".../mathlib.c/CopySubmatrix(): the submatrix of x2_dm must be within the ranges of both x1_dm and x2_dm");
}
+
#if defined( INTELCMATHLIBRARY )
void CopySubrowmatrix(TSdmatrix *x1_dm, const int br1, const int bc1, TSdmatrix *x2_dm, const int br2, const int bc2, const int nrs, const int ncs)
{
diff --git a/CFiles/mathlib.h b/CFiles/mathlib.h
old mode 100755
new mode 100644
index fc5371872d9194965cfded5414a9e124ca28430c..ae0a099faa613e2cec2da79206aafd6e5953e0d7
--- a/CFiles/mathlib.h
+++ b/CFiles/mathlib.h
@@ -1,6 +1,6 @@
#ifndef __MATHLIB_H__
#define __MATHLIB_H__
- #include "tzmatlab.h"
+ #include "tzmatlab_dw.h"
#include "fn_filesetup.h" //Used to call WriteMatrix(FPTR_DEBUG,....).
//------------------------------------------------------
@@ -339,6 +339,7 @@
void CopyCell0(TSdcell *x1_dc, TSdcell *x2_dc);
void CopySubmatrix0(TSdmatrix *x1_dm, TSdmatrix *x2_dm, const int br, const int bc, const int nrs, const int ncs);
void CopySubmatrix(TSdmatrix *x1_dm, const int br1, const int bc1, TSdmatrix *x2_dm, const int br2, const int bc2, const int nrs, const int ncs);
+ void CopySubmatrixSelectedCols0(TSdmatrix *x1_dm, TSdmatrix *x2_dm, const int br, const int nrs, TSivector *locs4cols);
void CopySubrowmatrix(TSdmatrix *x1_dm, const int br1, const int bc1, TSdmatrix *x2_dm, const int br2, const int bc2, const int nrs, const int ncs);
//??????? NOT tested yet.
void CopySubmatrix2rowmatrix(TSdmatrix *x1_dm, const int br1, const int bc1, TSdmatrix *x2_dm, const int br2, const int bc2, const int nrs, const int ncs);
diff --git a/CFiles/numgradgenTemplate.c b/CFiles/numgradgenTemplate.c
deleted file mode 100755
index 31130c371f6b98b0fdcdb4e9ece11b99c71642bd..0000000000000000000000000000000000000000
--- a/CFiles/numgradgenTemplate.c
+++ /dev/null
@@ -1,83 +0,0 @@
-/* For the general purpose numerical gradient.This source file (function) must be *manually* copied to a specific subdirectory (another source file).
- * Before copying this file, there are a few lines that must be *manually* changed, marked by CCCCCCCCC:
- * (1) The .h file to be included.
- * (2) T_tag in the typedef.
- * (3) T_tag in the declaration of the structure.
-*/
-#include "swz_comfuns.h" //CCCCCCCCC.
-
-
-typedef double TFgradfcn(struct T_tag *); //CCCCCCCCC.
-//-------------------------------
-// Modified from gradcd_gen() in cstz.c for the general purpose.
-//-------------------------------
-#define STPS 1.0e-04 // 6.0554544523933391e-6 step size = pow(DBL_EPSILON,1.0/3)
-void numgrad_cd(void *Snotype_ps)
-{
- /* The no-type structure Snotype_ps must contain the output and input arguments with the *exact* names. */
- //Outputs:
- // grad_dv: the gradient n-by-1 g (no need to be initialized).
- //Inputs:
- // xg_dv: the n-by-1 vector point at which the gradient is evaluated. No change in the end although will be added or substracted by dh during the function (but in the end the original value will be put back).
- // fcn_grad(struct T_tag *): the function for which the gradient is evaluated
- // grdh: step size. If 0.0, then dh is set automatically; otherwise, grdh is taken as a step size, often set as 1.0e-004.
- // fatx: the value of (*fcn_grad)(xg_dv). NOT used in this function except dealing with the boundary (NEARINFINITY) for the
- // minimization problem, but to be compatible with a genral function call where, say, a forward-difference method or cubic
- // interpolation of central difference method will use f0.
-
- //double *g, double *x, int n, double (*fcn)(double *x, int n), double *grdh, double f0)
-
- //--- Essential elments from the structure S_ps.
- struct T_tag *S_ps = (T_tag *)Snotype_ps; //CCCCCCCCC.
- int n = S_ps->grad_dv->n;
- double *g = S_ps->grad_dv->v;
- double *x = S_ps->xg_dv->v;
- TFgradfcn *fcn = S_ps->fcn_grad;
- double grdh = S_ps->grdh;
- double f0 = S_ps->fatx; //At the original point x.
- //--- Local to this function only.
- int i;
- double dh, dhi, dh2i, fp, fm, tmp, *xp;
-
-
- if (grdh != 0.0) {
- dh2i = (dhi=1.0/(dh=grdh))/2.0;
- for (i=0, xp=x; i<n; i++, xp++, g++) {
- tmp = *xp;
- *xp += dh;
- //The following statement is bad because dh does not get reset at the beginning of the loop and thus may get changed continually within the loop.
- // dh = *xp - tmp; // This increases the precision slightly.
- fp = fcn(S_ps); //At the point x+dh.
- *xp = tmp - dh;
- fm = fcn(S_ps); //At the point x-dh.
-
- //=== Checking the boundary condition for the minimization problem.
- if (fp >= NEARINFINITY) *g = (f0-fm)*dhi;
- else if (fm >= NEARINFINITY) *g = (fp-f0)*dhi;
- else *g = (fp-fm)*dh2i;
-
- *xp = tmp; // Put the original value of x[i] back to x[i] so that the content x[i] is still unaltered.
- }
-
- }
- else {
- for (i=0, xp=x; i<n; i++, xp++, g++) {
- dh = fabs(*xp)<=1 ? STPS : STPS*(*xp);
- tmp = *xp;
- *xp += dh;
- dh = *xp - tmp; // This increases the precision slightly.
- fp = fcn(S_ps); //At the point x+dh.
- *xp = tmp - dh;
- fm = fcn(S_ps); //At the point x-dh.
-
- //=== Checking the boundary condition for the minimization problem.
- if (fp >= NEARINFINITY) *g = (f0-fm)/dh;
- else if (fm >= NEARINFINITY) *g = (fp-f0)/dh;
- else *g = (fp-fm)/(2.0*dh);
-
- *xp = tmp; // Put the original value of x[i] back to x[i] so that the content x[i] is still unaltered.
- }
- }
-}
-#undef STPS
-
diff --git a/CFiles/optpackage.c b/CFiles/optpackage.c
old mode 100755
new mode 100644
index e52975edd0cbdae066e4ec4a242eb149671ac724..1da1647b0dbe5e7a189045b69bd3a87d038c4c77
--- a/CFiles/optpackage.c
+++ b/CFiles/optpackage.c
@@ -25,6 +25,17 @@ static void gradcd_imslconlin(int n, double *x, double *g);
static double ObjFuncForModel_congrad(double *x0_p, int d_x0);
+// NPSOL related
+static TSdvector *XNPSOL_DV = NULL; //To save the minimized value in case the IMSL quits with a higher value.
+void confun(int *mode, int *ncnln, int *n, int *ldJ, int *needc, double *x, double *c, double *cJac, int *nstate);
+void npsol(int *n, int *nclin, int *ncnln, int *ldA, int *ldJ, int *ldR, double *A, double *bl, double *bu,
+ void (*funcon)(int *mode, int *ncnln, int *n, int *ldJ, int *needc, double *x, double *c, double *cJac, int *nstate),
+ void (*funobj)(int *mode, int *n, double *x, double *f, double *g, int *nstate),
+ int *inform, int *iter, int *istate, double *c, double *cJac, double *clamda, double *f, double *g, double *R, double *x,
+ int *iw, int *leniw, double *w, int *lenw);
+static void ObjFuncForModel_npsolconlin(int *mode, int *N, double *x, double *f, double *g, int *nstate);
+static struct TStateModel_tag *SetModelGlobalForNPSOLconlin(struct TStateModel_tag *smodel_ps);
+static void npsolconlin_SetPrintFile(char *filename);
////TSminpack *CreateTSminpack(TFminpackage *minfinder_func, TFminobj *minobj_func, TFmingrad *mingrad_func, TFSetPrintFile *printinterresults_func, const int n, const int package) //, const int indxAnag)
////TSminpack *CreateTSminpack(TFminfinder *minfinder_func, TFminobj *minobj_func, TFmingrad *mingrad_func, char *filename_printout, const int n, const int package) //, const int indxAnag)
@@ -108,6 +119,7 @@ TSminpack *DestroyTSminpack(TSminpack *minpack_ps)
static TSetc_csminwel *CreateTSetc_csminwel(FILE *fptr_input1, const int n, const int q, const int k)
{
//If fptr_input1==NULL or no no values supplied when fptr_input1 != NULL, default values are taken.
+
int _i;
//===
TSetc_csminwel *etc_csminwel_ps = tzMalloc(1, TSetc_csminwel);
@@ -401,7 +413,6 @@ void InitializeForMinproblem(struct TSminpack_tag *minpack_ps, char *filename_sp
}
/**/
//------- Step 3. -------
-#if defined (NEWVERSIONofDW_SWITCH)
void minfinder_blockcsminwel(struct TSminpack_tag *minpack_ps, int indx_findMLE)
{
//Better version (November 2007)
@@ -462,8 +473,7 @@ void minfinder_blockcsminwel(struct TSminpack_tag *minpack_ps, int indx_findMLE)
//---- Refreshing the parameters outside this function. TZ October 2007.
SetupObjectiveFunction(smodel_ps, x1_pd, x2_pd, x_dv->v);
- //--- minpack_ps->fret0 has to be initialized using InitializeForMinproblem() in the main function.
- logvalue = -( minpack_ps->fret = PosteriorObjectiveFunction(x_dv->v, x_dv->n) ); //Refreshing. logPosterirPdf. DW function.
+ logvalue = -( minpack_ps->fret0 = minpack_ps->fret = PosteriorObjectiveFunction(x_dv->v, x_dv->n) ); //Refreshing. logPosterirPdf. DW function.
fprintf(FPTR_OPT, "\n=========== Beginning Blockwise and Overall csminwel Minimizations =======================\nLog Peak Value: %.16e\n", logvalue);
fflush(FPTR_OPT);
@@ -598,7 +608,7 @@ void minfinder_blockcsminwel(struct TSminpack_tag *minpack_ps, int indx_findMLE)
DestroyMatrix_lf(H2_dm);
DestroyMatrix_lf(H_dm);
}
-#endif
+
//-----------------------------------------------------
// Minimization csminwel for the constant parameter model only. 5/24/04.
@@ -801,17 +811,22 @@ void minfinder_noblockimslconlin(struct TSpackage_imslconlin_tag *package_imslco
CopyVector0(package_imslconlin_ps->xsaved_dv, XIMSL_DV);
//+
SetModelGlobalForIMSLconlin(smodel_ps);
- if (imsl_d_min_con_gen_lin(ObjFuncForModel_imslconlin, x_dv->n, package_imslconlin_ps->ncons, package_imslconlin_ps->neqs,
- package_imslconlin_ps->ncons ? package_imslconlin_ps->lh_coefs_dv->v : NULL,
- package_imslconlin_ps->ncons ? package_imslconlin_ps->rh_constraints_dv->v : NULL,
- package_imslconlin_ps->lowbounds_dv->v, package_imslconlin_ps->upperbounds_dv->v,
- IMSL_XGUESS, xguess_dv->v, IMSL_GRADIENT, gradcd_imslconlin,
- IMSL_MAX_FCN, package_imslconlin_ps->itmax, IMSL_OBJ, &minpack_ps->fret,
- IMSL_TOLERANCE, package_imslconlin_ps->crit, IMSL_RETURN_USER, x_dv->v, 0))
- {
- printf("\nFinished: IMSL linearly-constrained optimization is successfully finished with the value of obj. fun.: %.16e\n", minpack_ps->fret);
- }
- else printf("\nWarning: IMSL linearly-constrained optimization fails, so the results from csminwel and congramin are used.\n");
+
+
+ //////////////////////////////////////////////////////////////////////////////////////////////////////////////
+ //////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+ //if (imsl_d_min_con_gen_lin(ObjFuncForModel_imslconlin, x_dv->n, package_imslconlin_ps->ncons, package_imslconlin_ps->neqs,
+ // package_imslconlin_ps->lh_coefs_dv->v,
+ // package_imslconlin_ps->rh_constraints_dv->v,
+ // package_imslconlin_ps->lowbounds_dv->v, package_imslconlin_ps->upperbounds_dv->v,
+ // IMSL_XGUESS, xguess_dv->v, IMSL_GRADIENT, gradcd_imslconlin,
+ // IMSL_MAX_FCN, package_imslconlin_ps->itmax, IMSL_OBJ, &minpack_ps->fret,
+ // IMSL_TOLERANCE, package_imslconlin_ps->crit, IMSL_RETURN_USER, x_dv->v, 0))
+ //{
+ // printf("\nFinished: IMSL linearly-constrained optimization is successfully finished with the value of obj. fun.: %.16e\n", minpack_ps->fret);
+ //}
+ //else printf("\nWarning: IMSL linearly-constrained optimization fails, so the results from csminwel and congramin are used.\n");
printf("\n===Ending the IMSL constrained optimization===\n");
//=== Printing out messages indicating that IMSL has bugs.
@@ -1107,3 +1122,375 @@ static struct TSetc_imslconlin_tag *CreateTSetc_imslconlin(struct TSminpack_tag
}
/**/
+
+
+//-----------------------------------------------------------------------
+// Using Linearly-constrained NPSOL minimization package.
+// Added: 4/21/10 - Jake Smith
+//-----------------------------------------------------------------------
+void minfinder_noblocknpsolconlin(struct TSpackage_npsolconlin_tag *package_npsolconlin_ps, struct TSminpack_tag *minpack_ps, char *filename_printout, int ntheta)
+{
+#ifdef _NPSOL_
+
+
+ //ntheta: number of free model parameters (NOT including free transition matrix Q parameters).
+ //filename_printout: the file that stores the intermediate results.
+
+ int i, j;
+
+ //--- Model or project specific structure.
+ struct TStateModel_tag *smodel_ps = ((struct TSetc_minproj_tag *)minpack_ps->etc_project_ps)->smodel_ps;
+ //---
+
+
+ TSdvector *x_dv = minpack_ps->x_dv;
+ TSdvector *g_dv = minpack_ps->g_dv;
+ double *x1_pd=NULL, *x2_pd=NULL;
+ //===
+ TSdvector *xguess_dv = CreateVector_lf(x_dv->n);
+
+ x1_pd = x_dv->v; //In the constant parameter model, this will point to invalid,
+ x2_pd = x_dv->v + ntheta; // but will be taken care of automatically by DW's function ConvertFreeParametersToQ().
+
+ CopyVector0(xguess_dv, x_dv);
+
+ //======= NPSOL linearly-constrained optimization, which makes sure that the boundary condition is met.
+ npsolconlin_SetPrintFile(filename_printout); //Set the print-out file outputsp_min_tag.prn.
+ printf("\n\n======= Starting the NPSOL constrained optimization======= \n\n");
+ fflush(stdout);
+ //====== The following linearly-constrained minimization works well for this kind of model but has a bugger of returning a higher value of the objective function.
+ CopyVector0(XNPSOL_DV, x_dv); //This is absolutely necessary because once imsl_d_min_con_gen_lin() is called, x_dv will be
+ // changed before ObjFuncForModel_imslconlin() is evaluated. It is possible that x_dv is changed
+ // so much that bad objective is returned and thus XIMSL_DV would be bad from the start, thus
+ // giving 1.eE+3000 from beginning to end.
+ GLB_FVALMIN = -LogPosterior_StatesIntegratedOut(smodel_ps);
+ CopyVector0(package_npsolconlin_ps->xsaved_dv, XNPSOL_DV);
+ //+
+ SetModelGlobalForNPSOLconlin(smodel_ps);
+
+ ////////////////////////////////////////////////////////////////
+ // Set up NPSOL variables
+ ////////////////////////////////////////////////////////////////
+ int N = package_npsolconlin_ps->n;
+ int nlin = package_npsolconlin_ps->num_lin;
+ int nclin = package_npsolconlin_ps->num_nlin;
+ int nbnd = N + nlin + nclin;
+
+ double x[N];
+ double g[N];
+ //double *x = malloc(sizeof(double)*N);
+ //double *g = malloc(sizeof(double)*N);
+
+ int ldJ = nclin;
+ if (nclin <= 0)
+ ldJ = 1;
+
+ int ldR = N;
+ int ldA = nlin;
+
+ double A[50][50];
+
+ // convert row major definition to column major
+ if (nlin > 0) {
+ TSdmatrix *m = package_npsolconlin_ps->A;
+ for (i=0; i<N; i++) {
+ for (j=0; j<ldA; j++) {
+ //TODO: Talk to Tao about how he wants this done
+ //A[i][j] = m->M[j][i];
+ }
+ }
+ }
+ else
+ {
+ ldA = 1;
+ for (i=0; i<N; i++)
+ A[i][0] = 0;
+ }
+
+ double lb[150];
+ double ub[150];
+
+
+ // the first part of ub/lb are the general bound constraints
+ for (i=0; i<N;i++) {
+ lb[i] = package_npsolconlin_ps->lb->v[i];
+ ub[i] = package_npsolconlin_ps->ub->v[i];
+ }
+ // the second part of the ub/lb are the linear constraints;
+ for (i=N; i<N+nlin; i++) {
+ lb[i] = package_npsolconlin_ps->lin_lb->v[i-N];
+ ub[i] = package_npsolconlin_ps->lin_ub->v[i-N];
+ }
+ // the third part of the ub/lb are the nonlinear constraints;
+ for (i=N+nlin; i<nbnd; i++) {
+ lb[i] = package_npsolconlin_ps->nlin_lb->v[i-N-nlin];
+ ub[i] = package_npsolconlin_ps->nlin_ub->v[i-N-nlin];
+ }
+
+ double result;
+ double *c = NULL;
+ c = malloc(sizeof(double)*nclin);
+
+ int inform;
+
+ // workspace
+ double *R = malloc(ldR*N*sizeof(double));
+ int itera;
+ int istate[nbnd];
+ int leniw = 3*N + nlin + 2*nclin;
+ int lenw = 2*pow(N,2) + N*nlin + 2*N*nclin + 20*N + 11*nlin + 21*nclin;
+ int *iw = malloc(sizeof(int)*leniw);
+ double *w = malloc(sizeof(double)*lenw);
+ double *cJac = malloc(sizeof(double)*1);
+ double *clambda = malloc(sizeof(double)*nbnd);
+
+ ////////////////////////////////////////////////////////////////
+ // CALL NPSOL
+ ////////////////////////////////////////////////////////////////
+
+ npsol((int*) &N, &nlin, &nclin, &ldA, &ldJ, &ldR, *A, lb, ub, &confun, &ObjFuncForModel_npsolconlin, &inform,
+ &itera, istate, c, cJac, clambda, &result, g_dv->v, R, x_dv->v, iw, &leniw, w, &lenw);
+
+ minpack_ps->fret = result;
+
+ if (inform == 0) {
+ // The function completed successfullly
+ printf("\n===Ending the NPSOL constrained optimization===\n");
+ }
+ else if (inform == 1) {
+ // The final iterate of x satisfies the optimality condition to eh accuracy, but has yet to converge
+ printf("\n===Ending the NPSOL constrained optimization, DID NOT CONVERGE===\n");
+ } else {
+ // There was a problem
+ printf("\n-------NPSOL had a problem estimating ----------\n");
+ printf("Error Code: %d\n\n", (int) inform);
+ }
+
+
+ //=== Printing out messages indicating that IMSL has bugs.
+ if (minpack_ps->fret > GLB_FVALMIN)
+ {
+ //NPSOL linearly-constrained optimization returns a higher obj. func. (a bug).
+ printf("\n----------NPSOL linearly-constrained minimization finished but with a higher objective function value!----------\n");
+ printf("The improperly-returned value is %.10f while the lowest value of the objective function is %.16e.\n\n", minpack_ps->fret, GLB_FVALMIN);
+ fflush(stdout);
+ fprintf(FPTR_DEBUG, "\n----------NPSOL linearly-constrained minimization finished but with a higher objective function value!----------\n");
+ fprintf(FPTR_DEBUG, "The improperly-returned value is %.16e while the lowest value of the objective function is %.16e.\n\n", minpack_ps->fret, GLB_FVALMIN);
+ fflush(FPTR_DEBUG);
+ }
+
+ ConvertFreeParametersToQ(smodel_ps,x2_pd);
+ //DW's function, which takes care of the degenerate case where x2_pd points to an
+ // invalid place as in the constant parameter case.
+ ConvertFreeParametersToTheta(smodel_ps,x1_pd); //DW's function, which calls TZ's function. So essentially it's TZ's function.
+
+ //Saved the last best results in case the IMSL quits with a bug.
+ CopyVector0(package_npsolconlin_ps->xsaved_dv, XNPSOL_DV);
+
+
+ //===
+ DestroyVector_lf(xguess_dv);
+
+#endif
+}
+
+//-----------------------------------------------------------------------
+// ObjFuncForModel_npsolconlin():
+// Objective Function for linear constrained NPSOL problem
+//
+// mode: negative to exit NPSOL
+// N: dimensions of X
+// x0: parameters to maximize over
+// f: value of objective function evaluated at x0
+// g: gradient vector evaluated at x0: g(j) -> df/dx_j
+// nstate: can be ignored (allows for if the objective function needs
+// time to initiate itself, then nstate=1 the first time it is called
+//-----------------------------------------------------------------------
+static void ObjFuncForModel_npsolconlin(int *mode, int *N, double *x0, double *f, double *g, int *nstate)
+{
+ TSdvector x0_sdv;
+ //
+ FILE *fptr_startingpoint_vec = NULL;
+ static int ncnt_fevals = -1;
+
+ // printf("\n----- Entering the objective function. ------");
+ // fflush(stdout);
+ x0_sdv.v = x0;
+ x0_sdv.n = *N;
+ x0_sdv.flag = V_DEF;
+
+ *f = -opt_logOverallPosteriorKernal(SMODEL_PS, &x0_sdv);
+ if ( GLB_DISPLAY) {
+ printf("\nValue of objective function at the %dth evaluation: %.16e\n", ++ncnt_fevals, *f);
+ fflush(stdout);
+ }
+ if (*f < GLB_FVALMIN) {
+ //=== Resets GLB_FVALMIN at *fret_p and then prints the intermediate point to a file.
+ fptr_startingpoint_vec = tzFopen(filename_sp_vec_minproj,"w");
+ fprintf(fptr_startingpoint_vec, "================= Output from IMSC linear constrained optimization ====================\n");
+ fprintf(fptr_startingpoint_vec, "NPSOL: Value of objective miminization function at the %dth iteration: %.15f\n", ncnt_fevals, GLB_FVALMIN=*f);
+ fprintf(fptr_startingpoint_vec, "--------Restarting point---------\n");
+ WriteVector(fptr_startingpoint_vec, &x0_sdv, " %0.16e ");
+ CopyVector0(XNPSOL_DV, &x0_sdv); //Saved in case the IMSL quits with a bug.
+ //=== Must print this results because imsl_d_min_con_gen_lin() has a bug and quits with a higher value. The printed-out results in the debug file may be used for imsl_d_min_con_nonlin() to continue.
+ fprintf(FPTR_DEBUG, "\nNPSOL: Value of objective miminization function at the %dth iteration: %.15f\n", ncnt_fevals, GLB_FVALMIN=*f);
+ fprintf(FPTR_DEBUG, "--------Restarting point---------\n");
+ WriteVector(FPTR_DEBUG, &x0_sdv, " %0.16e ");
+ fflush(FPTR_DEBUG);
+ }
+
+ // printf("\n----- Leaving the objective function. ------\n");
+ // fflush(stdout);
+
+
+ // Deal with the gradient
+
+ // printf("\n=== Entering the gradient function. ===\n");
+ // fflush(stdout);
+ // TODO: JAKE FIX
+ //GLB_DISPLAY = 0; //This guarantees that the objective function printouts will not show when the gradient is computed.
+ gradcd_gen(g, x0, *N, ObjFuncForModel_congrad, (double *)NULL, ObjFuncForModel_congrad(x0, *N));
+
+ //=== Prints the intermediate gradient to a file.
+ // fptr_startingpoint_vec = tzFopen(filename_sp_vec_minproj,"r");
+ // fprintf(fptr_startingpoint_vec, "--------Numerical gradient---------\n");
+ // for (_i=0; _i<n; _i++) fprintf(fptr_startingpoint_vec, " %0.16e ", g[_i]);
+ // tzFclose(fptr_startingpoint_vec);
+
+ GLB_DISPLAY = 1;
+ // printf("\n=== Leaving the gradient function. ===\n");
+ // fflush(stdout);
+
+ tzFclose(fptr_startingpoint_vec);
+
+}
+
+// Holder for eventual nonlinear constraints
+void confun(int *mode, int *ncnln, int *n, int *ldJ, int *needc, double *x, double *c, double *cJac, int *nstate)
+{
+ *mode=1;
+}
+
+//-----------------------------------------------------------------------
+// Linearly-constrained IMSL minimization package.
+//-----------------------------------------------------------------------
+struct TSpackage_npsolconlin_tag *CreateTSpackagae_npsolconlin(const int npars_tot, const int num_lin, const int num_nlin)
+{
+ // n: total number of variables
+ // num_lin: the number of linear constraints
+ // num_nlin: the number of non-linear constraints
+ //
+ // General bounds again for equality lb(i) = ub(i)
+ // lb: n x 1 lower bound on paramters
+ // ub: n x 1 upper bound on paramters
+ //
+ // LINEAR BOUNDS if you want equality on an equation, set
+ // lin_lb(i) == lin_ub(i)
+ // lin_lb: (num_lin x 1) the lower bound of linear constraints
+ // lin_ub: (num_lin x 1) the upper bound of linear constraints
+ //
+ // NON-LINEAR BOUNDS
+ // nlin_lb: (num_nlin x 1) the lower bound for nonlinear constraint
+ // nlin_ub: (num_nlin x 1) the upper bound for nonlinear constraint
+ //
+ // A: (num_nlin x n) linear constraint matrix, with each row representing an equation
+ //
+ //
+ // EX:
+ // n = 4;
+ // x_1 + x_2 + x_3 >= 5
+ // x_2 + 2*x_4 = 4
+ // ==>
+ // A[0][0] = 1;
+ // A[0][1] = 1;
+ // A[0][2] = 1;
+ // A[1][1] = 1;
+ // A[1][3] = 2;
+ // nlin_lb[0] = 5;
+ // nlin_lb[1] = 4;
+ // nlin_ub[1] = 4;
+
+ //===
+ struct TSpackage_npsolconlin_tag *package_npsolconlin_ps = tzMalloc(1, struct TSpackage_npsolconlin_tag);
+
+ if (npars_tot<=0)
+ fn_DisplayError("CreateTSpackage_npsolconlin(): make sure (1) # of equality constraints no greater than total # of constraints"
+ "\t and (2) number of free parameters must be greater than 0");
+ package_npsolconlin_ps->n = npars_tot;
+ package_npsolconlin_ps->num_lin = num_lin;
+ package_npsolconlin_ps->num_nlin = num_nlin;
+
+ if (num_lin<=0)
+ {
+ package_npsolconlin_ps->lin_lb = NULL;
+ package_npsolconlin_ps->lin_ub = NULL;
+ package_npsolconlin_ps->A = NULL;
+ }
+ else
+ {
+ package_npsolconlin_ps->lin_lb = CreateConstantVector_lf(num_lin, -NPSOLINF);
+ package_npsolconlin_ps->lin_ub = CreateConstantVector_lf(num_lin, NPSOLINF);
+ package_npsolconlin_ps->A = CreateConstantMatrix_lf(num_lin, npars_tot, 0);
+ }
+ if (num_nlin<=0)
+ {
+ package_npsolconlin_ps->nlin_lb = NULL;
+ package_npsolconlin_ps->nlin_ub = NULL;
+ }
+ else
+ {
+ package_npsolconlin_ps->nlin_lb = CreateConstantVector_lf(num_nlin, -NPSOLINF);
+ package_npsolconlin_ps->nlin_ub = CreateConstantVector_lf(num_nlin, NPSOLINF);
+ }
+
+ // default to blank non linear constraint function
+ package_npsolconlin_ps->nlin_func = &confun;
+
+
+ package_npsolconlin_ps->lb = CreateConstantVector_lf(npars_tot, -NPSOLINF);
+ package_npsolconlin_ps->ub = CreateConstantVector_lf(npars_tot, NPSOLINF);
+ //-
+ package_npsolconlin_ps->xsaved_dv = CreateVector_lf(package_npsolconlin_ps->n);
+ XNPSOL_DV = CreateVector_lf(package_npsolconlin_ps->n); //Used in ObjFuncForModel_npsolconlin() to save the minimized value in case the IMSL quits with a higher value.
+
+ package_npsolconlin_ps->crit = CRIT_NPSOLCONLIN;
+ package_npsolconlin_ps->itmax = ITMAX_NPSOLCONLIN;
+
+
+ return (package_npsolconlin_ps);
+}
+//---
+struct TSpackage_npsolconlin_tag *DestroyTSpackagae_npsolconlin(struct TSpackage_npsolconlin_tag *package_npsolconlin_ps)
+{
+ if (package_npsolconlin_ps)
+ {
+ DestroyVector_lf(package_npsolconlin_ps->nlin_lb);
+ DestroyVector_lf(package_npsolconlin_ps->nlin_ub);
+ DestroyVector_lf(package_npsolconlin_ps->lin_lb);
+ DestroyVector_lf(package_npsolconlin_ps->lin_ub);
+ DestroyMatrix_lf(package_npsolconlin_ps->A);
+
+ //
+ DestroyVector_lf(package_npsolconlin_ps->xsaved_dv);
+ DestroyVector_lf(XNPSOL_DV);
+
+ //===
+ free(package_npsolconlin_ps);
+ return ((struct TSpackage_npsolconlin_tag *)NULL);
+ }
+ else return (package_npsolconlin_ps);
+}
+
+static struct TStateModel_tag *SetModelGlobalForNPSOLconlin(struct TStateModel_tag *smodel_ps)
+{
+ //Returns the old pointer in order to preserve the previous value.
+ struct TStateModel_tag *tmp_ps = SMODEL_PS;
+ SMODEL_PS = smodel_ps;
+ return (tmp_ps);
+}
+
+static void npsolconlin_SetPrintFile(char *filename) {
+ if (!filename) sprintf(filename_sp_vec_minproj, "outdata5npsolonlin.txt"); //Default filename.
+ else if (STRLEN-1 < strlen(filename)) fn_DisplayError(".../optpackage.c: the allocated length STRLEN for filename_sp_vec_minproj is too short. Must increase the string length");
+ else strcpy(filename_sp_vec_minproj, filename);
+}
diff --git a/CFiles/optpackage.h b/CFiles/optpackage.h
old mode 100755
new mode 100644
index 5215390889707fd8a04ca50061050a93a4113a11..50d187fd7da8df10be17031d1a1199378f672aff
--- a/CFiles/optpackage.h
+++ b/CFiles/optpackage.h
@@ -20,9 +20,15 @@
#include "congradmin.h"
#include "fn_filesetup.h" //fn_SetFilePosition(), etc.
#include "mathlib.h" //CopyVector0(), etc.
-#include "switch_opt.h" //DW's optimization routines for Markov-switching models.
+#include "dw_switch_opt.h" //DW's optimization routines for Markov-switching models.
#include "cstz.h" //Used for gradcd_gen() only in the IMSL linear constrainted problem.
+// ARE WE GOING TO BE USING NPSOL?
+#ifdef _NPSOL_
+ #define npsol npsol_
+ #define npoptn npoptn_
+#endif
+
//-------------- Attributes for selecting optimization packages. --------------
#define MIN_DEFAULT 0 //0 or NULL: default or no minimization package.
#define MIN_CSMINWEL 0x0001 //1: csminwel unconstrained minimization package.
@@ -53,6 +59,15 @@
#define CRIT_IMSLCONLIN 1.0e-09 //Overall convergence criterion on the first-order conditions.
#define ITMAX_IMSLCONLIN 100000 //Maximum number of iterations.
+
+//-------------- Minimization package: linearly-nconstrained NPSOL. --------------
+#define CRIT_NPSOLCONLIN 1.0e-09 //Overall convergence criterion on the first-order conditions.
+#define ITMAX_NPSOLCONLIN 100000 //Maximum number of iterations.
+#define NPSOLINF 1.0e21
+#define MAXPARAMETERS 50;
+#define MAXLINCONSTRAINTS 20;
+
+
//-------------- Minimization package: conjugate gradient method I. --------------
#define CRIT_CONGRAD1 1.0e-09 //Overall convergence criterion on the first-order conditions.
#define ITMAX_CONGRAD1 100000 //Maximum number of iterations.
@@ -242,10 +257,9 @@ struct TSetc_minproj_tag *DestroyTSetc_minproj(struct TSetc_minproj_tag *);
//---
//void InitializeForMinproblem(struct TSminpack_tag *minpack_ps, char *filename_sp, TSdvector *gphi_dv, int indxStartValuesForMin);
//=== Step 3.
-#if defined (NEWVERSIONofDW_SWITCH)
void minfinder_blockcsminwel(struct TSminpack_tag *minpack_ps, int indx_findMLE); //Blockwise minimization.
//indx_findMLE: 1: find MLE without a prior, 0: find posterior (with a prior).
-#endif
+
@@ -339,6 +353,70 @@ typedef struct TSminpack_csminwel_tag {
+//================ For NPSOL multivariate linearly-constrained minimizaiton package only. ================//
+//-----------------------------------------------------------------------
+// Using Linearly-constrained NPSOL minimization package.
+// Added: 4/21/10 - Jake Smith
+//-----------------------------------------------------------------------
+
+typedef struct TSpackage_npsolconlin_tag
+{
+ // n: total number of variables
+ // nlin: the number of linear constraints
+ // nclin: the number of non-linear constraints
+ //
+ // General bounds again for equality lb(i) = ub(i)
+ // lb: n x 1 lower bound on paramters
+ // ub: n x 1 upper bound on paramters
+ //
+ // LINEAR BOUNDS if you want equality on an equation, set
+ // nlin_lb(i) == nlin_ub(i)
+ // nlin_lb: (nlin x 1) the lower bound of linear constraints
+ // nlin_ub: (nlin x 1) the upper bound of linear constraints
+ //
+ // NON-LINEAR BOUNDS
+ // nclin_lb: (nclin x 1) the lower bound for nonlinear constraint
+ // nclin_ub: (nclin x 1) the upper bound for nonlinear constraint
+ //
+ // A: (nclin x n) linear constraint matrix, with each row representing an equation
+
+ int n; //Total number of free parameters for the optimaization.
+ //For example, model free parameters + free transition matrix parameters in the regime-switching case.
+ int num_lin; //the number of linear constraints
+ int num_nlin; //the number of non-linear constraints
+
+ TSdmatrix *A; //(nclin x n) linear constraint matrix, with each row representing an equation
+
+ TSdvector *lb; // lb: n x 1 lower bound on paramters
+ TSdvector *ub; // ub: n x 1 upper bound on paramters
+ //
+ // LINEAR BOUNDS if you want equality on an equation, set
+ // lin_lb(i) == lin_ub(i)
+ TSdvector *lin_lb; // nlin_lb: (nlin x 1) the lower bound of linear constraints
+ TSdvector *lin_ub; // nlin_ub: (nlin x 1) the upper bound of linear constraints
+ //
+ // NON-LINEAR BOUNDS
+ TSdvector *nlin_lb; // nclin_lb: (nclin x 1) the lower bound for nonlinear constraint
+ TSdvector *nlin_ub; // nclin_ub: (nclin x 1) the upper bound for nonlinear constraint
+
+ //=== Other output.
+ TSdvector *xsaved_dv; //npars_tot-by-1. Saved the parameters that give the minimal value of the objective function.
+
+ // NONLINEAR CONSTRAINT FUNCTION
+ void *nlin_func;
+
+ //=== Other inputs.
+ double crit; //Overall convergence criterion on the first-order conditions.
+ int itmax; //Maximum number of iterations.
+} TSpackage_npsolconlin;
+//+
+struct TSpackage_npsolconlin_tag *CreateTSpackagae_npsolconlin(const int npars_tot, const int neqs, const int ncons);
+struct TSpackage_npsolconlin_tag *DestroyTSpackagae_npsolconlin(struct TSpackage_npsolconlin_tag *package_npsolconlin_ps);
+
+void minfinder_noblocknpsolconlin(struct TSpackage_npsolconlin_tag *package_npsolconlin_ps, struct TSminpack_tag *minpack_ps, char *filename_printout, int ntheta);
+static void ObjFuncForModel_npsolconlin(int *mode, int *N, double *x, double *f, double *g, int *nstate);
+
+
#endif
diff --git a/CFiles/rand.c b/CFiles/rand.c
old mode 100755
new mode 100644
index 397fbbe9f1667149aa6f3bf40d411c100c185af7..03d08db9f4ff01cc219237fe38e9fa0da7bcec46
--- a/CFiles/rand.c
+++ b/CFiles/rand.c
@@ -143,12 +143,15 @@ double gaussrnd(void)
When a<1.0, Johnk's generator (Devroye, page 418).
When a>1.0, a rejection method or Best's algorithm (Devroye, page 410).
- A general gamma variate can be obtained as follows. Let z=b*x. Then,
+ A general gamma variate can be obtained as follows. Let z=x/b. Then,
z is drawn from a general gamma distribution whose density is
- 1
- p(z) = -------------- z^(a-1) exp(-z/b).
- gamma(a) b^a
+ b^a
+ p(z) = ---------- z^(a-1) exp(-b*z).
+ gamma(a)
+
+ This density is the same as Matlab where b is an inverse scale parameter (same
+ as Gelman's book but different from Matlab).
Uses algorithm translated by Iskander Karibzhanov from the Matlab function
gamrnd.m, which follows Johnk's generator in Devroye ("Non-Uniform Random
@@ -158,16 +161,14 @@ double gammrnd(double a) {
double b = a-1.0,
u, v, w, x, y, z;
- if (a <= 0.0) {
+ if (a <= 0.0)
+ {
//** When used with a C MEX-file for MATLAB.
printf("\nThe input argument x for gammrnd(x) in rand.c must be a positive double.");
- #ifdef WIN_MATLABAPI
- mexErrMsgTxt("Error! ???."); // This terminates the entire Matlab program.
- #else
- //@ When used entirely with C.
- getchar();
- exit(1); // This exits the entire C program.
- #endif
+
+ //@ When used entirely with C.
+ getchar();
+ exit(1); // This exits the entire C program.
}
if (a==1.0)
return -log(unirnd());
@@ -198,6 +199,24 @@ double gammrnd(double a) {
}
+/*
+ Returns a random draw from beta(a,b). The density function is
+
+ gamma(a+b)
+ p(x) = ------------------- x^(a-1) (1-x)^(b-1).
+ gamma(a)*gamma(b)
+
+ where a>0, b>0 and gamma denotes the gamma function, which is defined as
+ the integral with respect to t from 0 to infinity of exp(-t)*t^(z-1).
+*/
+double betarnd(double a, double b)
+{
+ double xa=2.0*gammrnd(a),
+ xb=2.0*gammrnd(b);
+
+ return (xa/(xa+xb));
+}
+
/*
Returns the integral from -infinity to x of 1/sqrt(2*PI)*exp(-y^2/2).
@@ -281,6 +300,14 @@ double StandardNormalDouble(void) {
#endif
}
+double NormalDouble(double m, double s)
+{
+ //m: mean;
+ //s: standard deviation.
+
+ return (m+s*StandardNormalDouble());
+}
+
double LogNormalDouble(double mean, double std)
{
//Return x where log(x) is normal(mean, std^2).
@@ -297,19 +324,30 @@ double LogNormalDouble(double mean, double std)
double GammaDouble(const double a, const double b) {
- //The probability density of x is of the form: ( x^(a-1) exp(-x/b) )/( Gamma(a) * b^a ).
- //The same form as in the MATLAB (not Gelman et al's) notation.
+ //The probability density of x is of the form: ( x^(a-1) exp(-bx) )/( b^a /Gamma(a) ).
+ //The same form as in the Gelman et al's (not MATLAB) notation.
#ifdef IMSL_RANDOMNUMBERGENERATOR // IMSL optimization routines.
//The GFSR algorithm to generate random numbers.
double x;
imsls_d_random_gamma(1, a, IMSLS_RETURN_USER, &x, 0);
- if (b != 1.0) x *= b;
+ if (b != 1.0) x /= b;
return x;
#else CASE2_RANDOMNUMBERGENERATOR //Imported from the C recipe book -- Case 2 and my own (Iskander) code for generating a gamma distribution.
- return ( (b==1.0) ? gammrnd(a) : b*gammrnd(a) );
+ return ( (b==1.0) ? gammrnd(a) : gammrnd(a)/b );
#endif
}
+double InverseGammaDouble(const double a, const double b)
+{
+ //p(x) = ( b^a/Gamma(a) ) x^(-a-1) exp(-b/x) for a>0 and b>0.
+ // where a is shape and b is scale parameter.
+ //E(x) = b/(a-1) for a>1; var(x) = b^2/( (a-1)^2*(a-2) ) for a>2;
+ //Noninformative distribution: a,b -> 0.
+ //How to draw: (1) draw z from Gamma(a,b); (2) let x=1/z.
+
+ return (1.0/GammaDouble(a,b));
+}
+
void UniformVector(TSdvector *x_dv)
{
diff --git a/CFiles/rand.h b/CFiles/rand.h
index a12c28e9df8ec4c1c6d173e53411817571994986..01c3ab08d0063c403420678cd200eec41274f9f3 100755
--- a/CFiles/rand.h
+++ b/CFiles/rand.h
@@ -13,6 +13,7 @@
double unirnd(void);
double gaussrnd(void);
double gammrnd(double a);
+ double betarnd(double a, double b);
double cumulative_normal(double x);
@@ -32,8 +33,10 @@
void SetSeedTableGFSR_IMSL(int iseed, int *itable);
double UniformDouble(void);
double StandardNormalDouble(void);
+ double NormalDouble(double m, double s);
double LogNormalDouble(double mean, double std);
double GammaDouble(const double a, const double b);
+ double InverseGammaDouble(const double a, const double b);
double ChisquareDouble(const double v);
void UniformVector(TSdvector *x_dv);
void StandardNormalVector(TSdvector *x_dv);
diff --git a/CFiles/template_proj.vcproj b/CFiles/template_proj.vcproj
deleted file mode 100755
index ab4cb30ae2fedebea4bdd82ab1d6cc9e2633fa67..0000000000000000000000000000000000000000
--- a/CFiles/template_proj.vcproj
+++ /dev/null
@@ -1,376 +0,0 @@
-<?xml version="1.0" encoding="Windows-1252"?>
-<VisualStudioProject
- ProjectType="Visual C++"
- Version="8.00"
- Name="template_proj"
- ProjectGUID="{30EA1610-6051-4681-A29C-9D29389FE60F}"
- Keyword="Win32Proj"
- >
- <Platforms>
- <Platform
- Name="Win32"
- />
- </Platforms>
- <ToolFiles>
- </ToolFiles>
- <Configurations>
- <Configuration
- Name="Debug|Win32"
- OutputDirectory="Debug"
- IntermediateDirectory="Debug"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="0"
- PreprocessorDefinitions="WIN32;_DEBUG;_CONSOLE"
- MinimalRebuild="true"
- BasicRuntimeChecks="3"
- RuntimeLibrary="1"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="true"
- DebugInformationFormat="4"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="2"
- IgnoreAllDefaultLibraries="false"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib;libcmtd.lib"
- GenerateDebugInformation="true"
- SubSystem="1"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCWebDeploymentTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- <Configuration
- Name="Release|Win32"
- OutputDirectory="Release"
- IntermediateDirectory="Release"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="2"
- InlineFunctionExpansion="2"
- EnableIntrinsicFunctions="true"
- FavorSizeOrSpeed="1"
- OmitFramePointers="true"
- EnableFiberSafeOptimizations="false"
- WholeProgramOptimization="false"
- PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE"
- StringPooling="true"
- RuntimeLibrary="0"
- EnableFunctionLevelLinking="true"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="true"
- DebugInformationFormat="3"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="1"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib"
- GenerateDebugInformation="false"
- SubSystem="1"
- OptimizeReferences="2"
- EnableCOMDATFolding="2"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCWebDeploymentTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- </Configurations>
- <References>
- </References>
- <Files>
- <Filter
- Name="Source Files"
- Filter="cpp;c;cxx;def;odl;idl;hpj;bat;asm"
- >
- <File
- RelativePath="..\..\..\TZCcode\csminwel.c"
- >
- <FileConfiguration
- Name="Debug|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- <FileConfiguration
- Name="Release|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- </File>
- <File
- RelativePath="..\..\..\TZCcode\cstz.c"
- >
- <FileConfiguration
- Name="Debug|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- <FileConfiguration
- Name="Release|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- </File>
- <File
- RelativePath="..\..\..\TZCcode\DWMatrixCode\switch\estimate.c"
- >
- </File>
- <File
- RelativePath="..\..\..\TZCcode\DWMatrixCode\extern_dw_switch_var_est.c"
- >
- </File>
- <File
- RelativePath="..\..\..\TZCcode\fn_filesetup.c"
- >
- <FileConfiguration
- Name="Debug|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- <FileConfiguration
- Name="Release|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- </File>
- <File
- RelativePath="..\..\..\TZCcode\mathlib.c"
- >
- <FileConfiguration
- Name="Debug|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- <FileConfiguration
- Name="Release|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- </File>
- <File
- RelativePath="..\..\..\TZCcode\rand.c"
- >
- <FileConfiguration
- Name="Debug|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- <FileConfiguration
- Name="Release|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- </File>
- <File
- RelativePath="..\..\..\TZCcode\tzmatlab.c"
- >
- <FileConfiguration
- Name="Debug|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- <FileConfiguration
- Name="Release|Win32"
- >
- <Tool
- Name="VCCLCompilerTool"
- ObjectFile="$(IntDir)/$(InputName)1.obj"
- />
- </FileConfiguration>
- </File>
- </Filter>
- <Filter
- Name="Header Files"
- Filter="h;hpp;hxx;hm;inl;inc"
- >
- <File
- RelativePath=".\resource.h"
- >
- </File>
- </Filter>
- <Filter
- Name="Resource Files"
- Filter="rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe"
- >
- </Filter>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_iblas.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_scalar.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat_iblas.lib"
- >
- </File>
- <File
- RelativePath="..\..\..\..\Program Files\Microsoft Visual Studio 8\VC\lib\libcmt.lib"
- >
- </File>
- <File
- RelativePath="..\..\Program Files\Microsoft Visual Studio 8\VC\lib\libcmtd.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\8.1\ia32\lib\libguide40.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\8.1\ia32\lib\mkl_c.lib"
- >
- </File>
- </Files>
- <Globals>
- </Globals>
-</VisualStudioProject>
diff --git a/CFiles/template_proj2.vcproj b/CFiles/template_proj2.vcproj
deleted file mode 100755
index 4f4a33c93c4104784ee6ed71de382be5d61b9d52..0000000000000000000000000000000000000000
--- a/CFiles/template_proj2.vcproj
+++ /dev/null
@@ -1,291 +0,0 @@
-<?xml version="1.0" encoding="Windows-1252"?>
-<VisualStudioProject
- ProjectType="Visual C++"
- Version="9.00"
- Name="template_proj2"
- ProjectGUID="{30EA1610-6051-4681-A29C-9D29389FE60F}"
- Keyword="Win32Proj"
- TargetFrameworkVersion="131072"
- >
- <Platforms>
- <Platform
- Name="Win32"
- />
- </Platforms>
- <ToolFiles>
- </ToolFiles>
- <Configurations>
- <Configuration
- Name="Debug|Win32"
- OutputDirectory="Debug"
- IntermediateDirectory="Debug"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="0"
- PreprocessorDefinitions="WIN32;_DEBUG;_CONSOLE"
- MinimalRebuild="true"
- BasicRuntimeChecks="3"
- RuntimeLibrary="1"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="false"
- DebugInformationFormat="4"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="2"
- IgnoreAllDefaultLibraries="false"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib;libcmtd.lib"
- GenerateDebugInformation="true"
- SubSystem="1"
- RandomizedBaseAddress="1"
- DataExecutionPrevention="0"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- <Configuration
- Name="Release|Win32"
- OutputDirectory="Release"
- IntermediateDirectory="Release"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="2"
- InlineFunctionExpansion="2"
- EnableIntrinsicFunctions="true"
- FavorSizeOrSpeed="1"
- OmitFramePointers="true"
- EnableFiberSafeOptimizations="true"
- WholeProgramOptimization="false"
- PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE"
- StringPooling="true"
- RuntimeLibrary="0"
- EnableFunctionLevelLinking="true"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="false"
- DebugInformationFormat="3"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="1"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib"
- GenerateDebugInformation="false"
- SubSystem="1"
- OptimizeReferences="2"
- EnableCOMDATFolding="2"
- RandomizedBaseAddress="1"
- DataExecutionPrevention="0"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- </Configurations>
- <References>
- </References>
- <Files>
- <Filter
- Name="Source Files"
- Filter="cpp;c;cxx;def;odl;idl;hpj;bat;asm"
- >
- <File
- RelativePath="D:\ZhaData\TZCcode\congradmin.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\csminwel.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\cstz.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\DWMatrixCode\extern_dw_switch.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\fn_filesetup.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\gensys.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\gradcd_switch.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\kalman.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\mathlib.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\optpackage.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\rand.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\tzmatlab.c"
- >
- </File>
- </Filter>
- <Filter
- Name="Header Files"
- Filter="h;hpp;hxx;hm;inl;inc"
- >
- </Filter>
- <Filter
- Name="Resource Files"
- Filter="rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe"
- >
- </Filter>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_iblas.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_scalar.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat_iblas.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Microsoft Visual Studio 9.0\VC\lib\libcmt.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Microsoft Visual Studio 9.0\VC\lib\libcmtd.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\9.0\ia32\lib\libguide.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\9.0\ia32\lib\mkl_c.lib"
- >
- </File>
- </Files>
- <Globals>
- </Globals>
-</VisualStudioProject>
diff --git a/CFiles/template_proj2_MKL8.0.vcproj b/CFiles/template_proj2_MKL8.0.vcproj
deleted file mode 100755
index 14d59cce04c151f7b13b1042d4733e3e01b29e29..0000000000000000000000000000000000000000
--- a/CFiles/template_proj2_MKL8.0.vcproj
+++ /dev/null
@@ -1,296 +0,0 @@
-<?xml version="1.0" encoding="Windows-1252"?>
-<VisualStudioProject
- ProjectType="Visual C++"
- Version="8.00"
- Name="template_proj2"
- ProjectGUID="{30EA1610-6051-4681-A29C-9D29389FE60F}"
- Keyword="Win32Proj"
- >
- <Platforms>
- <Platform
- Name="Win32"
- />
- </Platforms>
- <ToolFiles>
- </ToolFiles>
- <Configurations>
- <Configuration
- Name="Debug|Win32"
- OutputDirectory="Debug"
- IntermediateDirectory="Debug"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="0"
- PreprocessorDefinitions="WIN32;_DEBUG;_CONSOLE"
- MinimalRebuild="true"
- BasicRuntimeChecks="3"
- RuntimeLibrary="1"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="true"
- DebugInformationFormat="4"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="2"
- IgnoreAllDefaultLibraries="false"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib;libcmtd.lib"
- GenerateDebugInformation="true"
- SubSystem="1"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCWebDeploymentTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- <Configuration
- Name="Release|Win32"
- OutputDirectory="Release"
- IntermediateDirectory="Release"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="2"
- InlineFunctionExpansion="2"
- EnableIntrinsicFunctions="true"
- FavorSizeOrSpeed="1"
- OmitFramePointers="true"
- EnableFiberSafeOptimizations="false"
- WholeProgramOptimization="false"
- PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE"
- StringPooling="true"
- RuntimeLibrary="0"
- EnableFunctionLevelLinking="true"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="true"
- DebugInformationFormat="3"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="1"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib"
- GenerateDebugInformation="false"
- SubSystem="1"
- OptimizeReferences="2"
- EnableCOMDATFolding="2"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCWebDeploymentTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- </Configurations>
- <References>
- </References>
- <Files>
- <Filter
- Name="Source Files"
- Filter="cpp;c;cxx;def;odl;idl;hpj;bat;asm"
- >
- <File
- RelativePath="D:\ZhaData\TZCcode\congradmin.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\csminwel.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\cstz.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\DWMatrixCode\extern_dw_switch.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\fn_filesetup.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\\TZCcode\gradcd_switch.c"
- >
- </File>
- <File
- RelativePath=".\hyper_est.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\\TZCcode\kalman.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\\TZCcode\mathlib.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\\TZCcode\optpackage.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\rand.c"
- >
- </File>
- <File
- RelativePath="..\swz2_comfuns.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\tzmatlab.c"
- >
- </File>
- </Filter>
- <Filter
- Name="Header Files"
- Filter="h;hpp;hxx;hm;inl;inc"
- >
- </Filter>
- <Filter
- Name="Resource Files"
- Filter="rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe"
- >
- </Filter>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_iblas.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_scalar.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat_iblas.lib"
- >
- </File>
- <File
- RelativePath="D:\Program Files\Microsoft Visual Studio 8\VC\lib\libcmt.lib"
- >
- </File>
- <File
- RelativePath="D:\Program Files\Microsoft Visual Studio 8\VC\lib\libcmtd.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\8.1\ia32\lib\libguide40.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\8.1\ia32\lib\mkl_c.lib"
- >
- </File>
- </Files>
- <Globals>
- </Globals>
-</VisualStudioProject>
diff --git a/CFiles/template_proj2_MSV2005_MKL9.0.vcproj b/CFiles/template_proj2_MSV2005_MKL9.0.vcproj
deleted file mode 100755
index 3b7f01c6aeb3a00a4450efb885627006a68f0c1e..0000000000000000000000000000000000000000
--- a/CFiles/template_proj2_MSV2005_MKL9.0.vcproj
+++ /dev/null
@@ -1,292 +0,0 @@
-<?xml version="1.0" encoding="Windows-1252"?>
-<VisualStudioProject
- ProjectType="Visual C++"
- Version="8.00"
- Name="template_proj2"
- ProjectGUID="{30EA1610-6051-4681-A29C-9D29389FE60F}"
- Keyword="Win32Proj"
- >
- <Platforms>
- <Platform
- Name="Win32"
- />
- </Platforms>
- <ToolFiles>
- </ToolFiles>
- <Configurations>
- <Configuration
- Name="Debug|Win32"
- OutputDirectory="Debug"
- IntermediateDirectory="Debug"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="0"
- PreprocessorDefinitions="WIN32;_DEBUG;_CONSOLE"
- MinimalRebuild="true"
- BasicRuntimeChecks="3"
- RuntimeLibrary="1"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="true"
- DebugInformationFormat="4"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="2"
- IgnoreAllDefaultLibraries="false"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib;libcmtd.lib"
- GenerateDebugInformation="true"
- SubSystem="1"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCWebDeploymentTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- <Configuration
- Name="Release|Win32"
- OutputDirectory="Release"
- IntermediateDirectory="Release"
- ConfigurationType="1"
- InheritedPropertySheets="$(VCInstallDir)VCProjectDefaults\UpgradeFromVC70.vsprops"
- CharacterSet="2"
- >
- <Tool
- Name="VCPreBuildEventTool"
- />
- <Tool
- Name="VCCustomBuildTool"
- />
- <Tool
- Name="VCXMLDataGeneratorTool"
- />
- <Tool
- Name="VCWebServiceProxyGeneratorTool"
- />
- <Tool
- Name="VCMIDLTool"
- />
- <Tool
- Name="VCCLCompilerTool"
- Optimization="2"
- InlineFunctionExpansion="2"
- EnableIntrinsicFunctions="true"
- FavorSizeOrSpeed="1"
- OmitFramePointers="true"
- EnableFiberSafeOptimizations="false"
- WholeProgramOptimization="false"
- PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE"
- StringPooling="true"
- RuntimeLibrary="0"
- EnableFunctionLevelLinking="true"
- UsePrecompiledHeader="0"
- WarningLevel="3"
- Detect64BitPortabilityProblems="true"
- DebugInformationFormat="3"
- CompileAs="1"
- DisableSpecificWarnings="4996"
- />
- <Tool
- Name="VCManagedResourceCompilerTool"
- />
- <Tool
- Name="VCResourceCompilerTool"
- />
- <Tool
- Name="VCPreLinkEventTool"
- />
- <Tool
- Name="VCLinkerTool"
- AdditionalDependencies="netapi32.lib advapi32.lib gdi32.lib comdlg32.lib comctl32.lib wsock32.lib"
- LinkIncremental="1"
- IgnoreDefaultLibraryNames="libcd.lib;libc.lib"
- GenerateDebugInformation="false"
- SubSystem="1"
- OptimizeReferences="2"
- EnableCOMDATFolding="2"
- TargetMachine="1"
- />
- <Tool
- Name="VCALinkTool"
- />
- <Tool
- Name="VCManifestTool"
- />
- <Tool
- Name="VCXDCMakeTool"
- />
- <Tool
- Name="VCBscMakeTool"
- />
- <Tool
- Name="VCFxCopTool"
- />
- <Tool
- Name="VCAppVerifierTool"
- />
- <Tool
- Name="VCWebDeploymentTool"
- />
- <Tool
- Name="VCPostBuildEventTool"
- />
- </Configuration>
- </Configurations>
- <References>
- </References>
- <Files>
- <Filter
- Name="Source Files"
- Filter="cpp;c;cxx;def;odl;idl;hpj;bat;asm"
- >
- <File
- RelativePath="D:\ZhaData\TZCcode\congradmin.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\csminwel.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\cstz.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\DWMatrixCode\extern_dw_switch.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\fn_filesetup.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\gradcd_switch.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\gensys.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\kalman.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\mathlib.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\optpackage.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\rand.c"
- >
- </File>
- <File
- RelativePath="D:\ZhaData\TZCcode\tzmatlab.c"
- >
- </File>
- </Filter>
- <Filter
- Name="Header Files"
- Filter="h;hpp;hxx;hm;inl;inc"
- >
- </Filter>
- <Filter
- Name="Resource Files"
- Filter="rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe"
- >
- </Filter>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_iblas.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcmath_scalar.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\VNI\imsl\cnl600\vs05pc\lib\imslcstat_iblas.lib"
- >
- </File>
- <File
- RelativePath="D:\Program Files\Microsoft Visual Studio 8\VC\lib\libcmt.lib"
- >
- </File>
- <File
- RelativePath="D:\Program Files\Microsoft Visual Studio 8\VC\lib\libcmtd.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\9.0\ia32\lib\libguide40.lib"
- >
- </File>
- <File
- RelativePath="C:\Program Files\Intel\MKL\9.0\ia32\lib\mkl_c.lib"
- >
- </File>
- </Files>
- <Globals>
- </Globals>
-</VisualStudioProject>
diff --git a/CFiles/tzmatlab.c b/CFiles/tzmatlab.c
old mode 100755
new mode 100644
index b6c59d424338fc495ad62d3206289156a2a849f0..299daea904ff1e9a922e236b8c0478fa67f319fb
--- a/CFiles/tzmatlab.c
+++ b/CFiles/tzmatlab.c
@@ -9,7 +9,7 @@
/**/
-#include "tzmatlab.h"
+#include "tzmatlab_dw.h"
FILE *FPTR_DEBUG = (FILE *)NULL; //Debug output file, to be opened by main.c.
@@ -21,12 +21,15 @@ FILE *FPTR_OPT = (FILE *)NULL; //Optimization output file, to be opened by main
int fn_locofyearqm(int q_m, int yrstart, int qmstart, int yrend, int qmend)
{
//Returns the (base 0) location of a specified year and month (quarter) for the time series.
- //All the other inputs take the usual (base-1) numbers, I guess 01/17/05. For example, yrstart = 1960 means the year 1960.
+ //All the other inputs take the usual (base-1) numbers. For example, yrstart = 1960 means the year 1960.
int tmpi, loc;
if ( q_m != 12 )
if ( q_m != 4 ) fn_DisplayError(".../tzmatlab.c/fn_locofyearqm(): This function only works for monthly or quarterly data");
+ if ( (qmstart>q_m) || (qmend>q_m) )
+ fn_DisplayError(".../tzmatlab.c/fn_locofyearqm(): quarters must be no greater than 4 and months must be no greater than 12");
+
if ( (tmpi=yrend - yrstart) < 0 )
fn_DisplayError(".../cstz.c/fn_locofyearqm(): the end year must be greater than or equal to the start year");
else if ( (loc = (tmpi==0) ? qmend-qmstart : tmpi*q_m+qmend-qmstart) < 0 )
@@ -541,7 +544,7 @@ TSdcell *CreateConstantCell_lf(TSivector *row_iv, TSivector *col_iv, const doubl
TSdvector *CreateDatesVector_lf(int nq_m, int yrstart, int qmstart, int yrend, int qmend)
{
//If nq_m==4, quarterly data; nq_m==12, monthly data.
- //All the other inputs take the usual (base-1) numbers, I guess 01/17/05. For example, yrstart = 1960 means the year 1960.
+ //All the other inputs take the usual (base-1) numbers. For example, yrstart = 1960 means the year 1960.
int _t;
int samplesize = 1+fn_locofyearqm(nq_m, yrstart, qmstart, yrend, qmend); //1+ because fn_locofyearqm() returns a 0-based integer.
//
@@ -549,6 +552,9 @@ TSdvector *CreateDatesVector_lf(int nq_m, int yrstart, int qmstart, int yrend, i
dates_dv->n = samplesize;
dates_dv->v = tzMalloc(samplesize, double);
+ if ( (qmstart>nq_m) || (qmend>nq_m) )
+ fn_DisplayError(".../tzmatlab.c/CreateDatesVector_lf(): quarters must be no greater than 4 and months must be no greater than 12");
+
if (nq_m==4 || nq_m==12) {
for (_t=samplesize-1; _t>=0; _t--) dates_dv->v[_t] = (double)yrstart + (double)(qmstart+_t-1)/(double)nq_m;
dates_dv->flag = V_DEF;
diff --git a/CFiles/tzmatlab.h b/CFiles/tzmatlab.h
old mode 100755
new mode 100644
index 5c7c2d285027289e91d09afecbd27804f93633eb..b319d3568a1b355156f3b5f012df5efebc1eef04
--- a/CFiles/tzmatlab.h
+++ b/CFiles/tzmatlab.h
@@ -48,12 +48,6 @@
#include <time.h> //For time(), etc.
-
- //#include "mkl_cblas.h" // Intel MKL C Blas library.
- #include "mkl.h" //This includes mkl_cblas.h and mkl_lapack.h.
- #include <imsl.h> //IMSL math package.
- #include <imsls.h> //IMSL statistical package.
-
#define USE_DEBUG_FILE //When defined, one must use tzFopen to give the file name in the main .c file.
extern FILE *FPTR_DEBUG; //For debugging. Applied to all functions and all .c files that call tzmatlab.h.
//Initiated to NULL in tzmatlab.c.
@@ -63,7 +57,16 @@
//Initiated to NULL in tzmatlab.c.
//Must use tzFopen to give the file name in the main .c file.
- #define NEWVERSIONofDW_SWITCH //If defined, using DW's new switch program (implemented in 2008),
+/*******************************************************************************/
+/* Added by DW 9/1/08 */
+/*******************************************************************************/
+//#define USE_IMSL_MATH_LIBRARY
+//#define USE_IMSL_STAT_LIBRARY
+#define USE_GSL_LIBRARY
+#define USE_MKL_LIBRARY
+/*******************************************************************************/
+
+#define NEWVERSIONofDW_SWITCH //If defined, using DW's new switch program (implemented in 2008),
// which may be incompatible with previous programs, such as ...\SargentWZ2\EstProbModel\EstimationJuly07USED
//If undef, using the old, working switch program for, say, ...\SargentWZ2\EstProbModel\EstimationJuly07USED.
//Files that are affected are: cstz.c, kalman.c, optpackage.c,
@@ -93,7 +96,41 @@
//-------Only one of the following statistical packages.--------
#define IMSL_STATISTICSTOOLBOX // IMSL statistical package.
+/*******************************************************************************/
+/* Added by DW 9/1/08 */
+/*******************************************************************************/
+#if defined(USE_MKL_LIBRARY)
+ #include "mkl.h"
+#else
+ #if defined (USE_GSL_LIBRARY)
+ #include "gsl_cblas.h"
+ #endif
+ #include "blas_lapack.h"
+ #undef SWITCHTOINTELCMATH
+ #undef INTELCMATHLIBRARY
+#endif
+#if defined(USE_GSL_LIBRARY)
+ #include "gsl_sf_gamma.h"
+ #include "gsl_cdf.h"
+#endif
+
+#if defined(USE_IMSL_MATH_LIBRARY)
+ #include <imsl.h> //IMSL math package.
+ #include <imsls.h> //IMSL statistical package.
+#else
+ #undef IMSL_OPTIMIZATION
+ #undef SWITCHTOIMSLCMATH
+ #undef IMSL_OPTIMIZATION
+ #undef IMSL_RANDOMNUMBERGENERATOR
+#endif
+
+#if defined(USE_IMSL_STAT_LIBRARY)
+ #include <imsls.h> //IMSL statistical package.
+#else
+ #undef IMSL_STATISTICSTOOLBOX
+#endif
+/*******************************************************************************/
//-------If define: use matlab API interface; otherwise (undef), use C console.
#undef WIN_MATLABAPI // define: use matlab API interface; undef: use C dos console.
diff --git a/CFiles/tzmatlab_dw.h b/CFiles/tzmatlab_dw.h
new file mode 100644
index 0000000000000000000000000000000000000000..0d2449ae007127c9da3503d1bf202a61c322ffea
--- /dev/null
+++ b/CFiles/tzmatlab_dw.h
@@ -0,0 +1,363 @@
+/*********
+ * _cv: Pointer to TScvector (character vector).
+ * _iv: Pointer to TSivector (integer vector).
+ * _im: Pointer to TSimatrix (integer matrix).
+ * _dv: Pointer to TSdvector (double vector).
+ * _dm: Pointer to TSdmatrix (double matrix).
+ * _dc: Pointer to TSdcell (double cell -- pointer to pointer to a matrix).
+ * _dcv: Pointer to TSdcellvec (double cell -- pointer to pointer to a vector).
+ * _d4: Pointer to TSdfourth (double fourth dimension -- pointer to pointer to pointer to a matrix).
+ * _dzv: Pointer to TSdzvector (double complex vector).
+ * _dzm: Pointer to TSdzmatrix (double complex matrix).
+ *
+ * _s: structure variable.
+ * _ps: pointer to a structure.
+ * _sv: an array of structures.
+ *
+ * _sdv: structure (NOT pointer to structure) that contains TSdvector.
+ * _sdm: structure (NOT pointer to structure) that contains TSdmatrix.
+ *
+ * ???????? OLD NOTATIONS ??????????
+ * _v: C row or column vector pointer.
+ * _vint: C row or column vector pointer to integer.
+ * _m: C matrix pointer.
+ * _mint: C matrix pointer to integer.
+ * _m3: C 3-D matrix pointer.
+ * _ppm: C pointer to pointer to a matrix.
+ * d_???_ppm: the number of pointers that are pointed to by _ppm.
+ * rv_???_ppm: a vector (with dimension d_???_ppm) pointer of numbers of rows, each of the numbers coresponding to a pointed matrix.
+ * cv_???_ppm: a vector (with dimension d_???_ppm) pointer of numbers of columns, each of the numbers coresponding to a pointed matrix.
+ * d_???_v: dimension size.
+ * r_???_m: numbers of rows.
+ * c_???_m: number of columns.
+ * r_???_m3: number of rows.
+ * c_???_m3: number of columns.
+ * t_???_m3: number of a third dimension.
+*********/
+
+
+#ifndef __TZMATLAB__
+#define __TZMATLAB__
+ #define _ISOC99_SOURCE //Using C99 features for gcc or icc on Linux. Must be placed as the first line above all #include lines.
+
+ #include <stdio.h>
+ #include <stdlib.h> // For rand(), size_t, exit, malloc, free, qsort, EXIT_FAILURE.
+ #include <memory.h> //For memcpy, etc. Alternative: string.h
+ #include <math.h> //For isfinite.
+ #include <float.h> //For DBL_MIN, etc.
+ #include <time.h> //For time(), etc.
+
+
+ #define USE_DEBUG_FILE //When defined, one must use tzFopen to give the file name in the main .c file.
+ extern FILE *FPTR_DEBUG; //For debugging. Applied to all functions and all .c files that call tzmatlab.h.
+ //Initiated to NULL in tzmatlab.c.
+ //Must use tzFopen to give the file name in the main .c file.
+ extern FILE *FPTR_OPT; //For recording the optimization intermediate results.
+ //Applied to minfinder_blockcsminwel() in optpackage.c.
+ //Initiated to NULL in tzmatlab.c.
+ //Must use tzFopen to give the file name in the main .c file.
+
+/*******************************************************************************/
+/* Added by DW 9/1/08 */
+/*******************************************************************************/
+//#define USE_IMSL_MATH_LIBRARY
+//#define USE_IMSL_STAT_LIBRARY
+#define USE_GSL_LIBRARY
+#define USE_MKL_LIBRARY
+/*******************************************************************************/
+
+// #define NEWVERSIONofDW_SWITCH //If defined, using DW's new switch program (implemented in 2008),
+ // which may be incompatible with previous programs, such as ...\SargentWZ2\EstProbModel\EstimationJuly07USED
+ //If undef, using the old, working switch program for, say, ...\SargentWZ2\EstProbModel\EstimationJuly07USED.
+ //Files that are affected are: cstz.c, kalman.c, optpackage.c,
+
+
+ #define SWITCHTOIMSLCMATH // define: use IMSL special functions like gammlog; undef: use my own default code if it exists.
+
+ //-------Only one of the following for math library.--------
+ #define INTELCMATHLIBRARY // define: use Intel MKL LAPACK library; undef: use others.
+ //#define IMSLCMATHLIBRARY // define: use IMSL C Math library; undef: use others.
+ //#define MATLABCMATHLIBRARY // define: use Matlab C math library; undef: use others.
+
+ //-------Only one of the following for math library.--------
+ #define SWITCHTOINTELCMATH // define: use Intel MKL LAPACK library; undef: use others.
+ //#define SWITCHTOTZCMATH // define: use my own C math library; undef: use others.
+
+ //-------Only one of the following for optimization routines except that CG?_ and CSMINWEL_ can be chosen together.--------
+ //#define IMSL_OPTIMIZATION // IMSL optimization routines.
+ #define CSMINWEL_OPTIMIZATION //Sims's optimization routine.
+ #define CGI_OPTIMIZATION //Polak-Ribiere conjugate gradient method without using derivative information in performing the line minimization.
+ //NOT available yet! #define CGII_OPTIMIZATION //NOT available yet! Pletcher-Reeves conjugate gradient method using derivative information in performing the line minimization.
+
+ //-------Only one of the following for random number generating routines.--------
+ #define IMSL_RANDOMNUMBERGENERATOR // IMSL random number generator.
+ //#define CASE2_RANDOMNUMBERGENERATOR //Imported from the C recipe book -- case 2 and my own (Iskander) code for generating a gamma distribution.
+
+ //-------Only one of the following statistical packages.--------
+ #define IMSL_STATISTICSTOOLBOX // IMSL statistical package.
+
+/*******************************************************************************/
+/* Added by DW 9/1/08 */
+/*******************************************************************************/
+#if defined(USE_MKL_LIBRARY)
+ #include "mkl.h"
+#else
+ #if defined (USE_GSL_LIBRARY)
+ #include "gsl_cblas.h"
+ #endif
+ #include "blas_lapack.h"
+ #undef SWITCHTOINTELCMATH
+ #undef INTELCMATHLIBRARY
+#endif
+
+#if defined(USE_GSL_LIBRARY)
+ #include "gsl_sf_gamma.h"
+ #include "gsl_cdf.h"
+#endif
+
+#if defined(USE_IMSL_MATH_LIBRARY)
+ #include <imsl.h> //IMSL math package.
+ #include <imsls.h> //IMSL statistical package.
+#else
+ #undef IMSL_OPTIMIZATION
+ #undef SWITCHTOIMSLCMATH
+ #undef IMSL_OPTIMIZATION
+ #undef IMSL_RANDOMNUMBERGENERATOR
+#endif
+
+#if defined(USE_IMSL_STAT_LIBRARY)
+ #include <imsls.h> //IMSL statistical package.
+#else
+ #undef IMSL_STATISTICSTOOLBOX
+#endif
+/*******************************************************************************/
+
+ //-------If define: use matlab API interface; otherwise (undef), use C console.
+ #undef WIN_MATLABAPI // define: use matlab API interface; undef: use C dos console.
+
+
+ //---------------
+ #ifdef MATLABCMATHLIBRARY
+ #include "matlab.h" // For all mlf???? functions.
+ #include "matrix.h" // For mxGetM, mxCreatDoubleMatrix, etc.
+ #endif
+ #ifdef WIN_MATLABAPI // define: use matlab API interface; undef: use C dos console.
+ #include "mex.h" // For all mex??? calls. Matlab API (application program interface or external interface).
+ #define printf mexPrintf
+ #define malloc mxMalloc
+ #define calloc mxCalloc
+ #define free mxFree
+ #endif
+
+
+ //-------------- Attributes for the real double matrix type TSdmatrix. --------------
+ //-------------- Whenever a matrix is initialized, the default is M_GE, but nothing else. --------------
+ #define M_UNDEF 0 //0 or NULL: No attribute will be given when memory is allocated but no values are initialized.
+ #define M_GE 0x0001 //1: A general matrix.
+ #define M_SU 0x0002 //2: A symmetric (must be square) matrix but only the upper triangular part is referenced.
+ #define M_SL 0x0004 //4: A symmetric (must be square) matrix but only the lower triangular part is referenced.
+ #define M_UT 0x0008 //8: A upper triangular (trapezoidal if nrows < ncols) matrix but only the upper triangular part is referenced.
+ #define M_LT 0x0010 //16: A lower triangular (trapezoidal if nrows > ncols) matrix but only the lower triangular part is referenced.
+ #define M_CN 0x0020 //32: A constant (CN) matrix (All elements are the same or no (N) change from one to another).
+// #define M_UTU 0x0040 //2^6: An unit upper triangular matrix.
+// #define M_LTU 0x0080 //2^7: An unit lower triangular matrix.
+ //-------------- Attributes for the real double vector type TSdvector or the character vector type TScvector. --------------
+ #define V_UNDEF 0 //Zero or NULL: No values have been assigned to the double vector.
+ #define V_DEF 1 //True: Values have been assigned to the double vector.
+
+
+ //-------------- Other macro definitions. --------------
+ #define BLOCKSIZE_FOR_INTEL_MKL 128 //A machine-dependent value (typically, 16 to 64) required for optimum performance of the blocked algorithm in Intel MKL.
+ #define NEARINFINITY 1.0E+300
+ #define BIGREALNUMBER 1.0E+30
+ #define MACHINEZERO DBL_MIN
+ #define EPSILON DBL_EPSILON //1.0E-015. In Standard C, DBL_EPSILON = 2.2204460492503131
+ #define SQRTEPSILON 1.490116119384766E-008 //1.0E-15. In Standard C, DBL_EPSILON = 2.2204460492503131E-016
+ #define SQRTMACHINEZERO 1.490116119384766E-008
+ //This is really not correct, because this number is sqrt(epsion), where DBL_MIN is around 1.0e-300.
+ #define MACHINEINFINITY DBL_MAX
+ #define MACHINE_EXP_INFINITY DBL_MAX_EXP
+ #define EXP_NEARINFINITY 1000
+ //===
+ #define TZ_TRUE 1
+ #define TZ_FALSE 0
+
+
+
+ //---------------
+ #define tzMalloc(elt_count, type) (type *)m_alloc((elt_count)*sizeof(type))
+ #define tzCalloc(elt_count, type) (type *)c_alloc((elt_count), sizeof(type))
+ #define tzDestroy(x) {if ((x)) { \
+ free((x)); \
+ (x) = NULL; \
+ }}
+ #define tzFclose(x) {if ((x)) { \
+ fclose((x)); \
+ (x) = (FILE *)NULL; \
+ }}
+ #define mos(i, j, nrows) ((j)*(nrows)+(i)) //i: ith row; j: jth column; nrows: number of rows for the matrix.
+ //Offset(os) for a matrix(m) in column major order and with base 0. See Reek pp.241-242.
+ #define square(x) ((x) * (x)) //Must be careful to avoid using, say, square(tmpd=2) or square(++x).
+ #define _max(a, b) ((a)>(b) ? (a) : (b)) // Macro max or __max is already defined in stdlib.h in MS visual C++, but mex.h may overwrite the max macro so we use _max.
+ #define _min(a, b) ((a)>(b) ? (b) : (a))
+ #define swap(a, b, stemp) {(stemp)=(a); (a)=(b); (b)=(stemp);}
+ //
+ #ifndef isfinite
+ #define isfinite(x) _finite(x) //_finite is for Microsoft C++ compiler only (in float.h, which strangely is not ANSI compible),
+ // All these Microsoft functions are not yet C99 compatible.
+ #endif
+ //--- The following does not work.
+ // #ifndef FP_NAN
+ // #define FP_NAN _FPCLASS_SNAN //_FPCLASS_SNAN is for Microsoft C++ compiler only (in float.h, which strangely is not ANSI compible),
+ // // All these Microsoft functions are not yet C99 compatible.
+ // #endif
+ #define isdiagonalmatrix(x) (((x)->flag & (M_UT | M_LT)) == (M_UT | M_LT)) //x is the tz type of matrix.
+ //
+ #define DestroyDatesVector_lf(x) DestroyVector_lf(x)
+
+
+ //---------------
+ typedef struct TScvector_tag
+ {
+ char *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TScvector;
+ typedef struct TSvoidvector_tag
+ {
+ void *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TSvoidvector;
+ typedef struct {
+ int *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TSivector;
+ typedef struct {
+ int *M; //M: matrix.
+ int nrows, ncols;
+ int flag; //flag: Refers to M_GE, M_SU, M_SL, M_UT, and M_LT in tzmatlab.h.
+ } TSimatrix;
+ typedef struct {
+ TSivector **C; //ncells-by-1 cells (C) and a ponter to vector in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSicellvec;
+ typedef struct {
+ TSimatrix **C; //ncells-by-1 cells (C) and a ponter to vector in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSicell;
+ //=== Real types.
+ typedef struct {
+ double *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TSdvector;
+ typedef struct {
+ double *M; //M: matrix.
+ int nrows, ncols;
+ int flag; //flag: Refers to M_GE, M_SU, M_SL, M_UT, and M_LT in tzmatlab.h.
+ } TSdmatrix;
+ typedef struct {
+ TSdmatrix **C; //ncells-by-1 cells (C) and a pointer to matrix in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSdcell;
+ typedef struct {
+ TSdvector **C; //ncells-by-1 cells (C) and a ponter to vector in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSdcellvec;
+ typedef struct {
+ TSdcell **F; //ndims-by-1 fourths (F) and a pointer to cell in each fourth.
+ int ndims; //Number of pointers (fourth dimensions) to pointer.
+ } TSdfourth;
+ typedef struct {
+ TSdcellvec **F; //ndims-by-1 fourths (F) and a pointer to cellvec in each fourth.
+ int ndims; //Number of pointers (fourth dimensions) to pointer.
+ } TSdfourthvec;
+ //=== Complex types.
+ typedef struct {
+ TSdvector *real; //Real part.
+ TSdvector *imag; //Imaginary part.
+ } TSdzvector;
+ typedef struct {
+ TSdmatrix *real; //Real part.
+ TSdmatrix *imag; //Imaginary part.
+ } TSdzmatrix;
+
+
+
+ //----------------- Some high-level functions. -----------------
+ int fn_locofyearqm(int q_m, int yrstart, int qmstart, int yrend, int qmend);
+
+
+
+
+ //---------------
+ void fn_DisplayError(char *msg_s);
+ void *m_alloc(size_t size);
+ void *c_alloc(size_t elt_count, size_t elt_size);
+
+ /**
+ TSvoidvector *CreateVector_void(int _n);
+ TSvoidvector *DestroyVector_void(TSvoidvector *x_voidv);
+ /**/
+
+ TScvector *CreateVector_c(int _n);
+ TScvector *DestroyVector_c(TScvector *x_s);
+ TSivector *CreateVector_int(int _n);
+ TSivector *DestroyVector_int(TSivector *x_iv);
+ TSimatrix *CreateMatrix_int(int nrows, int ncols);
+ TSimatrix *DestroyMatrix_int(TSimatrix *x_im);
+ TSicellvec *CreateCellvec_int(TSivector *n_iv);
+ TSicellvec *DestroyCellvec_int(TSicellvec *x_icv);
+ TSicell *CreateCell_int(TSivector *row_iv, TSivector *col_iv);
+ TSicell *DestroyCell_int(TSicell *x_ic);
+
+ TSdvector *CreateVector_lf(int _n);
+ TSdvector *DestroyVector_lf(TSdvector *x_iv);
+ TSdmatrix *CreateMatrix_lf(int nrows, int ncols);
+ TSdmatrix *DestroyMatrix_lf(TSdmatrix *x_im);
+ TSdcell *CreateCell_lf(TSivector *row_iv, TSivector *col_iv);
+ TSdcell *DestroyCell_lf(TSdcell *x_dc);
+ TSdcellvec *CreateCellvec_lf(TSivector *n_iv);
+ TSdcellvec *DestroyCellvec_lf(TSdcellvec *x_dcv);
+ TSdfourth *CreateFourth_lf(int ndims, TSivector *row_iv, TSivector *col_iv);
+ TSdfourth *DestroyFourth_lf(TSdfourth *x_d4);
+ TSdfourthvec *CreateFourthvec_lf(int ndims, TSivector *n_iv);
+ TSdfourthvec *DestroyFourthvec_lf(TSdfourthvec *x_d4v);
+
+ TSdzvector *CreateVector_dz(int _n);
+ TSdzvector *DestroyVector_dz(TSdzvector *x_dzv);
+ TSdzmatrix *CreateMatrix_dz(int nrows, int ncols);
+ TSdzmatrix *DestroyMatrix_dz(TSdzmatrix *x_dzm);
+
+ //+
+ TSdmatrix *CreateZeroMatrix_lf(const int nrows, const int ncols);
+ TSdmatrix *CreateIdentityMatrix_lf(const int nrows, const int ncols);
+ //TSdvector *CreateZerosVector_lf(int _n);
+ TSivector *CreateConstantVector_int( const int _n, const int _k);
+ TSimatrix *CreateConstantMatrix_int(const int nrows, const int ncols, const int _n);
+ TSicellvec *CreateConstantCellvec_int(TSivector *n_iv, const int _n);
+ TSicell *CreateConstantCell_int(TSivector *row_iv, TSivector *col_iv, const int _n);
+ TSdvector *CreateConstantVector_lf(const int _n, const double _alpha);
+ TSdmatrix *CreateConstantMatrix_lf(const int nrows, const int ncols, const double _alpha);
+ TSdcellvec *CreateConstantCellvec_lf(TSivector *n_iv, const double _alpha);
+ TSdcell *CreateConstantCell_lf(TSivector *row_iv, TSivector *col_iv, const double _alpha);
+ TSdvector *CreateDatesVector_lf(int nq_m, int yrstart, int qmstart, int yrend, int qmend);
+ //+
+ void InitializeConstantVector_lf(TSdvector *x_dv, const double _alpha);
+ void InitializeConstantVector_int(TSivector *x_iv, const int _k);
+ void InitializeConstantMatrix_lf(TSdmatrix *x_dm, const double _alpha);
+ void InitializeDiagonalMatrix_lf(TSdmatrix *x_dm, const double _alpha);
+ void InitializeConstantMatrix_int(TSimatrix *x_dm, const int _alpha);
+ void InitializeConstantCellvec_lf(TSdcellvec *x_dcv, const double _alpha);
+ void InitializeConstantCell_lf(TSdcell *x_dc, const double _alpha);
+ void InitializeConstantFourthvec_lf(TSdfourthvec *x_d4v, const double _alpha);
+ void InitializeConstantFourth_lf(TSdfourth *x_d4, const double _alpha);
+
+
+ void NegateColofMatrix_lf(TSdvector *y_dv, TSdmatrix *x_dm, int _j);
+ void InitializeConstantColofMatrix_lf(TSdmatrix *X_dm, int jx, double _alpha);
+
+ FILE *tzFopen(char *filename, char *mode);
+#endif
diff --git a/CFiles/tzmatlab_tao.h b/CFiles/tzmatlab_tao.h
new file mode 100644
index 0000000000000000000000000000000000000000..e8da1ac75ec89f031c686cafa8fdd03968c0ebaf
--- /dev/null
+++ b/CFiles/tzmatlab_tao.h
@@ -0,0 +1,359 @@
+/*********
+ * _cv: Pointer to TScvector (character vector).
+ * _iv: Pointer to TSivector (integer vector).
+ * _im: Pointer to TSimatrix (integer matrix).
+ * _dv: Pointer to TSdvector (double vector).
+ * _dm: Pointer to TSdmatrix (double matrix).
+ * _dc: Pointer to TSdcell (double cell -- pointer to pointer to a matrix).
+ * _dcv: Pointer to TSdcellvec (double cell -- pointer to pointer to a vector).
+ * _d4: Pointer to TSdfourth (double fourth dimension -- pointer to pointer to pointer to a matrix).
+ * _dzv: Pointer to TSdzvector (double complex vector).
+ * _dzm: Pointer to TSdzmatrix (double complex matrix).
+ *
+ * _s: structure variable.
+ * _ps: pointer to a structure.
+ * _sv: an array of structures.
+ *
+ * _sdv: structure (NOT pointer to structure) that contains TSdvector.
+ * _sdm: structure (NOT pointer to structure) that contains TSdmatrix.
+ *
+ * ???????? OLD NOTATIONS ??????????
+ * _v: C row or column vector pointer.
+ * _vint: C row or column vector pointer to integer.
+ * _m: C matrix pointer.
+ * _mint: C matrix pointer to integer.
+ * _m3: C 3-D matrix pointer.
+ * _ppm: C pointer to pointer to a matrix.
+ * d_???_ppm: the number of pointers that are pointed to by _ppm.
+ * rv_???_ppm: a vector (with dimension d_???_ppm) pointer of numbers of rows, each of the numbers coresponding to a pointed matrix.
+ * cv_???_ppm: a vector (with dimension d_???_ppm) pointer of numbers of columns, each of the numbers coresponding to a pointed matrix.
+ * d_???_v: dimension size.
+ * r_???_m: numbers of rows.
+ * c_???_m: number of columns.
+ * r_???_m3: number of rows.
+ * c_???_m3: number of columns.
+ * t_???_m3: number of a third dimension.
+*********/
+
+
+#ifndef __TZMATLAB__
+#define __TZMATLAB__
+ #define _ISOC99_SOURCE //Using C99 features for gcc or icc on Linux. Must be placed as the first line above all #include lines.
+
+ #include <stdio.h>
+ #include <stdlib.h> // For rand(), size_t, exit, malloc, free, qsort, EXIT_FAILURE.
+ #include <memory.h> //For memcpy, etc. Alternative: string.h
+ #include <math.h> //For isfinite.
+ #include <float.h> //For DBL_MIN, etc.
+ #include <time.h> //For time(), etc.
+
+
+ #define USE_DEBUG_FILE //When defined, one must use tzFopen to give the file name in the main .c file.
+ extern FILE *FPTR_DEBUG; //For debugging. Applied to all functions and all .c files that call tzmatlab.h.
+ //Initiated to NULL in tzmatlab.c.
+ //Must use tzFopen to give the file name in the main .c file.
+ extern FILE *FPTR_OPT; //For recording the optimization intermediate results.
+ //Applied to minfinder_blockcsminwel() in optpackage.c.
+ //Initiated to NULL in tzmatlab.c.
+ //Must use tzFopen to give the file name in the main .c file.
+
+/*******************************************************************************/
+/* Added by DW 9/1/08 */
+/*******************************************************************************/
+//#define USE_IMSL_MATH_LIBRARY
+//#define USE_IMSL_STAT_LIBRARY
+#define USE_GSL_LIBRARY
+#define USE_MKL_LIBRARY
+/*******************************************************************************/
+
+ #define NEWVERSIONofDW_SWITCH //If defined, using DW's new switch program (implemented in 2008),
+ // which may be incompatible with previous programs, such as ...\SargentWZ2\EstProbModel\EstimationJuly07USED
+ //If undef, using the old, working switch program for, say, ...\SargentWZ2\EstProbModel\EstimationJuly07USED.
+ //Files that are affected are: cstz.c, kalman.c, optpackage.c,
+
+
+ #define SWITCHTOIMSLCMATH // define: use IMSL special functions like gammlog; undef: use my own default code if it exists.
+
+ //-------Only one of the following for math library.--------
+ #define INTELCMATHLIBRARY // define: use Intel MKL LAPACK library; undef: use others.
+ //#define IMSLCMATHLIBRARY // define: use IMSL C Math library; undef: use others.
+ //#define MATLABCMATHLIBRARY // define: use Matlab C math library; undef: use others.
+
+ //-------Only one of the following for math library.--------
+ #define SWITCHTOINTELCMATH // define: use Intel MKL LAPACK library; undef: use others.
+ //#define SWITCHTOTZCMATH // define: use my own C math library; undef: use others.
+
+ //-------Only one of the following for optimization routines except that CG?_ and CSMINWEL_ can be chosen together.--------
+ //#define IMSL_OPTIMIZATION // IMSL optimization routines.
+ #define CSMINWEL_OPTIMIZATION //Sims's optimization routine.
+ #define CGI_OPTIMIZATION //Polak-Ribiere conjugate gradient method without using derivative information in performing the line minimization.
+ //NOT available yet! #define CGII_OPTIMIZATION //NOT available yet! Pletcher-Reeves conjugate gradient method using derivative information in performing the line minimization.
+
+ //-------Only one of the following for random number generating routines.--------
+ #define IMSL_RANDOMNUMBERGENERATOR // IMSL random number generator.
+ //#define CASE2_RANDOMNUMBERGENERATOR //Imported from the C recipe book -- case 2 and my own (Iskander) code for generating a gamma distribution.
+
+ //-------Only one of the following statistical packages.--------
+ #define IMSL_STATISTICSTOOLBOX // IMSL statistical package.
+
+/*******************************************************************************/
+/* Added by DW 9/1/08 */
+/*******************************************************************************/
+#if defined(USE_MKL_LIBRARY)
+ #include "mkl.h"
+#else
+ #include "blas_lapack.h"
+ #undef SWITCHTOINTELCMATH
+#endif
+
+#if defined(USE_GSL_LIBRARY)
+ #include "gsl_sf_gamma.h"
+ #include "gsl_cdf.h"
+#endif
+
+#if defined(USE_IMSL_MATH_LIBRARY)
+ #include <imsl.h> //IMSL math package.
+ #include <imsls.h> //IMSL statistical package.
+#else
+ #undef IMSL_OPTIMIZATION
+ #undef SWITCHTOIMSLCMATH
+ #undef IMSL_OPTIMIZATION
+ #undef IMSL_RANDOMNUMBERGENERATOR
+#endif
+
+#if defined(USE_IMSL_STAT_LIBRARY)
+ #include <imsls.h> //IMSL statistical package.
+#else
+ #undef IMSL_STATISTICSTOOLBOX
+#endif
+/*******************************************************************************/
+
+ //-------If define: use matlab API interface; otherwise (undef), use C console.
+ #undef WIN_MATLABAPI // define: use matlab API interface; undef: use C dos console.
+
+
+ //---------------
+ #ifdef MATLABCMATHLIBRARY
+ #include "matlab.h" // For all mlf???? functions.
+ #include "matrix.h" // For mxGetM, mxCreatDoubleMatrix, etc.
+ #endif
+ #ifdef WIN_MATLABAPI // define: use matlab API interface; undef: use C dos console.
+ #include "mex.h" // For all mex??? calls. Matlab API (application program interface or external interface).
+ #define printf mexPrintf
+ #define malloc mxMalloc
+ #define calloc mxCalloc
+ #define free mxFree
+ #endif
+
+
+ //-------------- Attributes for the real double matrix type TSdmatrix. --------------
+ //-------------- Whenever a matrix is initialized, the default is M_GE, but nothing else. --------------
+ #define M_UNDEF 0 //0 or NULL: No attribute will be given when memory is allocated but no values are initialized.
+ #define M_GE 0x0001 //1: A general matrix.
+ #define M_SU 0x0002 //2: A symmetric (must be square) matrix but only the upper triangular part is referenced.
+ #define M_SL 0x0004 //4: A symmetric (must be square) matrix but only the lower triangular part is referenced.
+ #define M_UT 0x0008 //8: A upper triangular (trapezoidal if nrows < ncols) matrix but only the upper triangular part is referenced.
+ #define M_LT 0x0010 //16: A lower triangular (trapezoidal if nrows > ncols) matrix but only the lower triangular part is referenced.
+ #define M_CN 0x0020 //32: A constant (CN) matrix (All elements are the same or no (N) change from one to another).
+// #define M_UTU 0x0040 //2^6: An unit upper triangular matrix.
+// #define M_LTU 0x0080 //2^7: An unit lower triangular matrix.
+ //-------------- Attributes for the real double vector type TSdvector or the character vector type TScvector. --------------
+ #define V_UNDEF 0 //Zero or NULL: No values have been assigned to the double vector.
+ #define V_DEF 1 //True: Values have been assigned to the double vector.
+
+
+ //-------------- Other macro definitions. --------------
+ #define BLOCKSIZE_FOR_INTEL_MKL 128 //A machine-dependent value (typically, 16 to 64) required for optimum performance of the blocked algorithm in Intel MKL.
+ #define NEARINFINITY 1.0E+300
+ #define BIGREALNUMBER 1.0E+30
+ #define MACHINEZERO DBL_MIN
+ #define EPSILON DBL_EPSILON //1.0E-015. In Standard C, DBL_EPSILON = 2.2204460492503131
+ #define SQRTEPSILON 1.490116119384766E-008 //1.0E-15. In Standard C, DBL_EPSILON = 2.2204460492503131E-016
+ #define SQRTMACHINEZERO 1.490116119384766E-008
+ //This is really not correct, because this number is sqrt(epsion), where DBL_MIN is around 1.0e-300.
+ #define MACHINEINFINITY DBL_MAX
+ #define MACHINE_EXP_INFINITY DBL_MAX_EXP
+ #define EXP_NEARINFINITY 1000
+ //===
+ #define TZ_TRUE 1
+ #define TZ_FALSE 0
+
+
+
+ //---------------
+ #define tzMalloc(elt_count, type) (type *)m_alloc((elt_count)*sizeof(type))
+ #define tzCalloc(elt_count, type) (type *)c_alloc((elt_count), sizeof(type))
+ #define tzDestroy(x) {if ((x)) { \
+ free((x)); \
+ (x) = NULL; \
+ }}
+ #define tzFclose(x) {if ((x)) { \
+ fclose((x)); \
+ (x) = (FILE *)NULL; \
+ }}
+ #define mos(i, j, nrows) ((j)*(nrows)+(i)) //i: ith row; j: jth column; nrows: number of rows for the matrix.
+ //Offset(os) for a matrix(m) in column major order and with base 0. See Reek pp.241-242.
+ #define square(x) ((x) * (x)) //Must be careful to avoid using, say, square(tmpd=2) or square(++x).
+ #define _max(a, b) ((a)>(b) ? (a) : (b)) // Macro max or __max is already defined in stdlib.h in MS visual C++, but mex.h may overwrite the max macro so we use _max.
+ #define _min(a, b) ((a)>(b) ? (b) : (a))
+ #define swap(a, b, stemp) {(stemp)=(a); (a)=(b); (b)=(stemp);}
+ //
+ #ifndef isfinite
+ #define isfinite(x) _finite(x) //_finite is for Microsoft C++ compiler only (in float.h, which strangely is not ANSI compible),
+ // All these Microsoft functions are not yet C99 compatible.
+ #endif
+ //--- The following does not work.
+ // #ifndef FP_NAN
+ // #define FP_NAN _FPCLASS_SNAN //_FPCLASS_SNAN is for Microsoft C++ compiler only (in float.h, which strangely is not ANSI compible),
+ // // All these Microsoft functions are not yet C99 compatible.
+ // #endif
+ #define isdiagonalmatrix(x) (((x)->flag & (M_UT | M_LT)) == (M_UT | M_LT)) //x is the tz type of matrix.
+ //
+ #define DestroyDatesVector_lf(x) DestroyVector_lf(x)
+
+
+ //---------------
+ typedef struct TScvector_tag
+ {
+ char *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TScvector;
+ typedef struct TSvoidvector_tag
+ {
+ void *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TSvoidvector;
+ typedef struct {
+ int *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TSivector;
+ typedef struct {
+ int *M; //M: matrix.
+ int nrows, ncols;
+ int flag; //flag: Refers to M_GE, M_SU, M_SL, M_UT, and M_LT in tzmatlab.h.
+ } TSimatrix;
+ typedef struct {
+ TSivector **C; //ncells-by-1 cells (C) and a ponter to vector in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSicellvec;
+ typedef struct {
+ TSimatrix **C; //ncells-by-1 cells (C) and a ponter to vector in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSicell;
+ //=== Real types.
+ typedef struct {
+ double *v; //v: vector.
+ int n;
+ int flag; //flag: no legal values are assigned if 0 and legal values are assigned if 1.
+ } TSdvector;
+ typedef struct {
+ double *M; //M: matrix.
+ int nrows, ncols;
+ int flag; //flag: Refers to M_GE, M_SU, M_SL, M_UT, and M_LT in tzmatlab.h.
+ } TSdmatrix;
+ typedef struct {
+ TSdmatrix **C; //ncells-by-1 cells (C) and a pointer to matrix in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSdcell;
+ typedef struct {
+ TSdvector **C; //ncells-by-1 cells (C) and a ponter to vector in each cell.
+ int ncells; //Number of pointers (cells) to pointer.
+ } TSdcellvec;
+ typedef struct {
+ TSdcell **F; //ndims-by-1 fourths (F) and a pointer to cell in each fourth.
+ int ndims; //Number of pointers (fourth dimensions) to pointer.
+ } TSdfourth;
+ typedef struct {
+ TSdcellvec **F; //ndims-by-1 fourths (F) and a pointer to cellvec in each fourth.
+ int ndims; //Number of pointers (fourth dimensions) to pointer.
+ } TSdfourthvec;
+ //=== Complex types.
+ typedef struct {
+ TSdvector *real; //Real part.
+ TSdvector *imag; //Imaginary part.
+ } TSdzvector;
+ typedef struct {
+ TSdmatrix *real; //Real part.
+ TSdmatrix *imag; //Imaginary part.
+ } TSdzmatrix;
+
+
+
+ //----------------- Some high-level functions. -----------------
+ int fn_locofyearqm(int q_m, int yrstart, int qmstart, int yrend, int qmend);
+
+
+
+
+ //---------------
+ void fn_DisplayError(char *msg_s);
+ void *m_alloc(size_t size);
+ void *c_alloc(size_t elt_count, size_t elt_size);
+
+ /**
+ TSvoidvector *CreateVector_void(int _n);
+ TSvoidvector *DestroyVector_void(TSvoidvector *x_voidv);
+ /**/
+
+ TScvector *CreateVector_c(int _n);
+ TScvector *DestroyVector_c(TScvector *x_s);
+ TSivector *CreateVector_int(int _n);
+ TSivector *DestroyVector_int(TSivector *x_iv);
+ TSimatrix *CreateMatrix_int(int nrows, int ncols);
+ TSimatrix *DestroyMatrix_int(TSimatrix *x_im);
+ TSicellvec *CreateCellvec_int(TSivector *n_iv);
+ TSicellvec *DestroyCellvec_int(TSicellvec *x_icv);
+ TSicell *CreateCell_int(TSivector *row_iv, TSivector *col_iv);
+ TSicell *DestroyCell_int(TSicell *x_ic);
+
+ TSdvector *CreateVector_lf(int _n);
+ TSdvector *DestroyVector_lf(TSdvector *x_iv);
+ TSdmatrix *CreateMatrix_lf(int nrows, int ncols);
+ TSdmatrix *DestroyMatrix_lf(TSdmatrix *x_im);
+ TSdcell *CreateCell_lf(TSivector *row_iv, TSivector *col_iv);
+ TSdcell *DestroyCell_lf(TSdcell *x_dc);
+ TSdcellvec *CreateCellvec_lf(TSivector *n_iv);
+ TSdcellvec *DestroyCellvec_lf(TSdcellvec *x_dcv);
+ TSdfourth *CreateFourth_lf(int ndims, TSivector *row_iv, TSivector *col_iv);
+ TSdfourth *DestroyFourth_lf(TSdfourth *x_d4);
+ TSdfourthvec *CreateFourthvec_lf(int ndims, TSivector *n_iv);
+ TSdfourthvec *DestroyFourthvec_lf(TSdfourthvec *x_d4v);
+
+ TSdzvector *CreateVector_dz(int _n);
+ TSdzvector *DestroyVector_dz(TSdzvector *x_dzv);
+ TSdzmatrix *CreateMatrix_dz(int nrows, int ncols);
+ TSdzmatrix *DestroyMatrix_dz(TSdzmatrix *x_dzm);
+
+ //+
+ TSdmatrix *CreateZeroMatrix_lf(const int nrows, const int ncols);
+ TSdmatrix *CreateIdentityMatrix_lf(const int nrows, const int ncols);
+ //TSdvector *CreateZerosVector_lf(int _n);
+ TSivector *CreateConstantVector_int( const int _n, const int _k);
+ TSimatrix *CreateConstantMatrix_int(const int nrows, const int ncols, const int _n);
+ TSicellvec *CreateConstantCellvec_int(TSivector *n_iv, const int _n);
+ TSicell *CreateConstantCell_int(TSivector *row_iv, TSivector *col_iv, const int _n);
+ TSdvector *CreateConstantVector_lf(const int _n, const double _alpha);
+ TSdmatrix *CreateConstantMatrix_lf(const int nrows, const int ncols, const double _alpha);
+ TSdcellvec *CreateConstantCellvec_lf(TSivector *n_iv, const double _alpha);
+ TSdcell *CreateConstantCell_lf(TSivector *row_iv, TSivector *col_iv, const double _alpha);
+ TSdvector *CreateDatesVector_lf(int nq_m, int yrstart, int qmstart, int yrend, int qmend);
+ //+
+ void InitializeConstantVector_lf(TSdvector *x_dv, const double _alpha);
+ void InitializeConstantVector_int(TSivector *x_iv, const int _k);
+ void InitializeConstantMatrix_lf(TSdmatrix *x_dm, const double _alpha);
+ void InitializeDiagonalMatrix_lf(TSdmatrix *x_dm, const double _alpha);
+ void InitializeConstantMatrix_int(TSimatrix *x_dm, const int _alpha);
+ void InitializeConstantCellvec_lf(TSdcellvec *x_dcv, const double _alpha);
+ void InitializeConstantCell_lf(TSdcell *x_dc, const double _alpha);
+ void InitializeConstantFourthvec_lf(TSdfourthvec *x_d4v, const double _alpha);
+ void InitializeConstantFourth_lf(TSdfourth *x_d4, const double _alpha);
+
+
+ void NegateColofMatrix_lf(TSdvector *y_dv, TSdmatrix *x_dm, int _j);
+ void InitializeConstantColofMatrix_lf(TSdmatrix *X_dm, int jx, double _alpha);
+
+ FILE *tzFopen(char *filename, char *mode);
+#endif
diff --git a/MatlabFiles/MSV/OldVersions/dw_gensys.m b/MatlabFiles/MSV/OldVersions/dw_gensys.m
new file mode 100755
index 0000000000000000000000000000000000000000..cd959321a68ebe04c981adf3bdaac052548533f2
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/dw_gensys.m
@@ -0,0 +1,46 @@
+function [G1,impact,gev,eu]=dw_gensys(g0,g1,psi,pi,div)
+% function [G1,impact,gev,eu]=gensys(g0,g1,c,psi,pi,div)
+% System given as
+% g0*y(t)=g1*y(t-1)+psi*e(t)+pi*eta(t),
+% with e an exogenous variable process and eta being endogenously determined
+% one-step-ahead expectational errors. Returned system is
+% y(t)=G1*y(t-1)+impact*e(t).
+% If div is omitted from argument list, a div>1 is calculated.
+% eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
+% existence only with not-s.c. z; eu=[-2,-2] for coincident zeros.
+% By Daniel Waggoner
+
+realsmall=1e-8;
+
+G1=[];
+impact=[];
+gev=[];
+
+n=size(pi,1);
+
+if nargin == 4
+ div=1.0;
+end
+
+[a b q z]=qz(g0,g1);
+
+for i=1:n
+ if (abs(a(i,i)) < realsmall) & (abs(b(i,i)) < realsmall)
+ disp('Coincident zeros.')
+ eu=[-2;-2];
+ gev=[diag(a) diag(b)];
+ return
+ end
+end
+
+[a b q z]=qzsort(a,b,q,z);
+
+suppress=0;
+i=n;
+while (i > 0) & (abs(b(i,i)) > abs(a(i,i)*div))
+ suppress=suppress+1;
+ i=i-1;
+end
+
+[G1,impact,eu]=qzcomputesolution(a,b,q,z,psi,pi,suppress);
+
diff --git a/MatlabFiles/MSV/OldVersions/gensys.m b/MatlabFiles/MSV/OldVersions/gensys.m
new file mode 100755
index 0000000000000000000000000000000000000000..69f7fc0fa0229de91d3d4cc9a2be063d24152c42
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/gensys.m
@@ -0,0 +1,214 @@
+function [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
+% function [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(g0,g1,c,psi,pi,div)
+% System given as
+% g0*y(t)=g1*y(t-1)+c+psi*z(t)+pi*eta(t),
+% with z an exogenous variable process and eta being endogenously determined
+% one-step-ahead expectational errors. Returned system is
+% y(t)=G1*y(t-1)+C+impact*z(t)+ywt*inv(I-fmat*inv(L))*fwt*z(t+1) .
+% If z(t) is i.i.d., the last term drops out.
+% If div is omitted from argument list, a div>1 is calculated.
+% eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
+% existence only with not-s.c. z; eu=[-2,-2] for coincident zeros.
+% By Christopher A. Sims
+% Corrected 10/28/96 by CAS
+
+eu=[0;0];
+realsmall=1e-7;
+fixdiv=(nargin==6);
+n=size(g0,1);
+[a b q z v]=qz(g0,g1);
+if ~fixdiv, div=1.01; end
+nunstab=0;
+zxz=0;
+for i=1:n
+% ------------------div calc------------
+ if ~fixdiv
+ if abs(a(i,i)) > 0
+ divhat=abs(b(i,i))/abs(a(i,i));
+ % bug detected by Vasco Curdia and Daria Finocchiaro, 2/25/2004 A root of
+ % exactly 1.01 and no root between 1 and 1.02, led to div being stuck at 1.01
+ % and the 1.01 root being misclassified as stable. Changing < to <= below fixes this.
+ if 1+realsmall<divhat & divhat<=div
+ div=.5*(1+divhat);
+ end
+ end
+ end
+% ----------------------------------------
+ nunstab=nunstab+(abs(b(i,i))>div*abs(a(i,i)));
+ if abs(a(i,i))<realsmall & abs(b(i,i))<realsmall
+ zxz=1;
+ end
+end
+div ;
+nunstab;
+if ~zxz
+ [a b q z]=qzdiv(div,a,b,q,z);
+end
+gev=[diag(a) diag(b)];
+if zxz
+ disp('Coincident zeros. Indeterminacy and/or nonexistence.')
+ eu=[-2;-2];
+ % correction added 7/29/2003. Otherwise the failure to set output
+ % arguments leads to an error message and no output (including eu).
+ G1=[];C=[];impact=[];fmat=[];fwt=[];ywt=[];gev=[];
+ return
+end
+q1=q(1:n-nunstab,:);
+q2=q(n-nunstab+1:n,:);
+z1=z(:,1:n-nunstab)';
+z2=z(:,n-nunstab+1:n)';
+a2=a(n-nunstab+1:n,n-nunstab+1:n);
+b2=b(n-nunstab+1:n,n-nunstab+1:n);
+etawt=q2*pi;
+% zwt=q2*psi;
+[ueta,deta,veta]=svd(etawt);
+md=min(size(deta));
+bigev=find(diag(deta(1:md,1:md))>realsmall);
+ueta=ueta(:,bigev);
+veta=veta(:,bigev);
+deta=deta(bigev,bigev);
+% ------ corrected code, 3/10/04
+eu(1) = length(bigev)>=nunstab;
+length(bigev);
+nunstab;
+% ------ Code below allowed "existence" in cases where the initial lagged state was free to take on values
+% ------ inconsistent with existence, so long as the state could w.p.1 remain consistent with a stable solution
+% ------ if its initial lagged value was consistent with a stable solution. This is a mistake, though perhaps there
+% ------ are situations where we would like to know that this "existence for restricted initial state" situation holds.
+%% [uz,dz,vz]=svd(zwt);
+%% md=min(size(dz));
+%% bigev=find(diag(dz(1:md,1:md))>realsmall);
+%% uz=uz(:,bigev);
+%% vz=vz(:,bigev);
+%% dz=dz(bigev,bigev);
+%% if isempty(bigev)
+%% exist=1;
+%% else
+%% exist=norm(uz-ueta*ueta'*uz) < realsmall*n;
+%% end
+%% if ~isempty(bigev)
+%% zwtx0=b2\zwt;
+%% zwtx=zwtx0;
+%% M=b2\a2;
+%% for i=2:nunstab
+%% zwtx=[M*zwtx zwtx0];
+%% end
+%% zwtx=b2*zwtx;
+%% [ux,dx,vx]=svd(zwtx);
+%% md=min(size(dx));
+%% bigev=find(diag(dx(1:md,1:md))>realsmall);
+%% ux=ux(:,bigev);
+%% vx=vx(:,bigev);
+%% dx=dx(bigev,bigev);
+%% existx=norm(ux-ueta*ueta'*ux) < realsmall*n;
+%% else
+%% existx=1;
+%% end
+% ----------------------------------------------------
+% Note that existence and uniqueness are not just matters of comparing
+% numbers of roots and numbers of endogenous errors. These counts are
+% reported below because usually they point to the source of the problem.
+% ------------------------------------------------------
+[ueta1,deta1,veta1]=svd(q1*pi);
+md=min(size(deta1));
+bigev=find(diag(deta1(1:md,1:md))>realsmall);
+ueta1=ueta1(:,bigev);
+veta1=veta1(:,bigev);
+deta1=deta1(bigev,bigev);
+%% if existx | nunstab==0
+%% %disp('solution exists');
+%% eu(1)=1;
+%% else
+%% if exist
+%% %disp('solution exists for unforecastable z only');
+%% eu(1)=-1;
+%% %else
+%% %fprintf(1,'No solution. %d unstable roots. %d endog errors.\n',nunstab,size(ueta1,2));
+%% end
+%% %disp('Generalized eigenvalues')
+%% %disp(gev);
+%% %md=abs(diag(a))>realsmall;
+%% %ev=diag(md.*diag(a)+(1-md).*diag(b))\ev;
+%% %disp(ev)
+%% % return;
+%% end
+if isempty(veta1)
+ unique=1;
+else
+ unique=norm(veta1-veta*veta'*veta1)<realsmall*n;
+end
+
+veta1;
+veta;
+norm(veta1-veta*veta'*veta1);
+
+if unique
+ %disp('solution unique');
+ eu(2)=1;
+else
+ fprintf(1,'Indeterminacy. %d loose endog errors.\n',size(veta1,2)-size(veta,2));
+ %disp('Generalized eigenvalues')
+ %disp(gev);
+ %md=abs(diag(a))>realsmall;
+ %ev=diag(md.*diag(a)+(1-md).*diag(b))\ev;
+ %disp(ev)
+% return;
+end
+tmat = [eye(n-nunstab) -(ueta*(deta\veta')*veta1*deta1*ueta1')'];
+G0= [tmat*a; zeros(nunstab,n-nunstab) eye(nunstab)];
+G1= [tmat*b; zeros(nunstab,n)];
+% ----------------------
+% G0 is always non-singular because by construction there are no zeros on
+% the diagonal of a(1:n-nunstab,1:n-nunstab), which forms G0's ul corner.
+% -----------------------
+G0I=inv(G0);
+G1=G0I*G1;
+usix=n-nunstab+1:n;
+C=G0I*[tmat*q*c;(a(usix,usix)-b(usix,usix))\q2*c];
+impact=G0I*[tmat*q*psi;zeros(nunstab,size(psi,2))];
+fmat=b(usix,usix)\a(usix,usix);
+fwt=-b(usix,usix)\q2*psi;
+ywt=G0I(:,usix);
+% -------------------- above are output for system in terms of z'y -------
+G1=real(z*G1*z');
+C=real(z*C);
+impact=real(z*impact);
+% Correction 10/28/96: formerly line below had real(z*ywt) on rhs, an error.
+ywt=z*ywt;
+
+%--------------------------------------------------------------------------
+% The code below check that a solution is obtained with stable manifold
+% z1. Should be commented out for production runs.
+%--------------------------------------------------------------------------
+% eu
+%
+% z2=z(:,n-nunstab+1:n);
+% z1=null(z2');
+%
+% n=norm(g0*(G1*z1) - g1*z1);
+% disp('norm(g0*(g1*z1) - g1*z1) -- should be zero');
+% disp(n);
+%
+% eta=-veta*diag(1./diag(deta))*ueta'*q2*psi;
+% n=norm(g0*impact - (psi + pi*eta));
+% disp('norm(g0*impact - psi - pi*eta) -- should be zero');
+% disp(n);
+%
+% n=norm(z2'*G1*z1);
+% disp('norm(z2''*G1*z1) -- should be zero');
+% disp(n);
+%
+% n=norm(z2'*G1);
+% disp('norm(z2''G1) -- best if zero');
+% disp(n);
+%
+% n=norm(z2'*impact);
+% disp('norm(z2''*impact) -- should be zero');
+% disp(n);
+%
+% n=norm(G1*z2);
+% disp('norm(G1*z2)');
+% disp(n);
+%
+% disp('press any key to continue');
+% pause
diff --git a/MatlabFiles/MSV/OldVersions/mainfunction.m b/MatlabFiles/MSV/OldVersions/mainfunction.m
new file mode 100755
index 0000000000000000000000000000000000000000..cf38e917c721f806fccd86e33fe94d6bb181ee4e
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/mainfunction.m
@@ -0,0 +1,219 @@
+% The constant paramater model written in the canonical form:
+% A x_t = B x_{t-1} + Gamma_u u_t + Pi pi_t
+% where f_t = [pi_t, y_t, R_t, mu_w_t, nu_t, E_t pi_{t+1}, E_t y_{t+1}]';
+% ut = [Sst*x_t{-1}; diag(e_t(1),...,e_t(h*))]
+% e_t is a 3-by-1 vector of fundamental shocks: e_{r t}, e_{wt}, and e_{nu t};
+% pi_t is a 2-by-1 vector of expectationsl errors for inflation and output: pi_{pi t} and pi_{y t}.
+%
+% Output format: x_t = G1*x_{t-1} + impact*e_t.
+%
+% Look for <<>> for ad hoc changes.
+%See Notes pp.41-42.
+
+smallval = 1.0e-8; %<sqrt(machine epsilon).
+locs_normalizedvars = [3 4 5]; %R_t, mu_w_t, nu_t;
+seednumber = 7910; %472534; % if 0, random state at each clock time
+if seednumber
+ randn('state',seednumber);
+ rand('state',seednumber);
+else
+ randn('state',fix(100*sum(clock)));
+ rand('state',fix(100*sum(clock)));
+end
+n_normalizedvars = length(locs_normalizedvars);
+
+
+
+%--- Reading in the values of the parameters.
+%addpath D:\ZhaData\WorkDisk\LiuWZ\CalibrationMatlab\LWZmodel\Paravals2r
+cd ./Paravals2r
+%paravals2r_bench
+%paravals2r_complex
+paravals2r_noexistence
+%paravals2r_try
+cd ..
+
+Q = [
+ q11 1-q22
+ 1-q11 q22
+ ];
+disp(' ')
+disp('Ergodic probability:')
+qbar = fn_ergodp(Q)
+g_etabar = qbar'*g_eta;
+if (flag_swap_rs)
+ g_eta = flipud(g_eta);
+ g_gamma = flipud(g_gamma);
+ g_rhor = flipud(g_rhor);
+ g_phipi = flipud(g_phipi);
+ g_phiy = flipud(g_phiy);
+ Q = flipud(fliplr(Q));
+end
+
+
+%--- Composite regime-switching parameters. The symbol "td" stands for "tilde."
+htd = 1; %4;
+Qtd = [
+ Q(1,1) Q(1,1) 0 0
+ 0 0 Q(1,2) Q(1,2)
+ Q(2,1) Q(2,1) 0 0
+ 0 0 Q(2,2) Q(2,2)
+ ];
+%--- psi1(std_t) = g_psi1(s_t,s_{t-1})
+psi1td = [
+ g_etabar*(1-g_eta(1)) / (g_eta(1)*(1-g_eta(1)))
+ g_etabar*(1-g_eta(2)) / (g_eta(2)*(1-g_eta(1)))
+ g_etabar*(1-g_eta(1)) / (g_eta(1)*(1-g_eta(2)))
+ g_etabar*(1-g_eta(2)) / (g_eta(2)*(1-g_eta(2)))
+ ];
+%--- psi2(std_t) = g_psi2(s_{t-1})
+psi2td = [
+ (1-g_beta*g_etabar)*(1-g_eta(1)) / (g_eta(1)*(1+g_thetap*(1-g_alpha)/g_alpha));
+ (1-g_beta*g_etabar)*(1-g_eta(2)) / (g_eta(2)*(1+g_thetap*(1-g_alpha)/g_alpha));
+ (1-g_beta*g_etabar)*(1-g_eta(1)) / (g_eta(1)*(1+g_thetap*(1-g_alpha)/g_alpha));
+ (1-g_beta*g_etabar)*(1-g_eta(2)) / (g_eta(2)*(1+g_thetap*(1-g_alpha)/g_alpha));
+ ];
+%--- gamma0(std_t) = g_gamma(s_{t})
+gamma0td = [
+ g_gamma(1)
+ g_gamma(1)
+ g_gamma(2)
+ g_gamma(2)
+ ];
+%--- gamma1(std_t) = g_gamma(s_{t-1})
+gamma1td = [
+ g_gamma(1)
+ g_gamma(2)
+ g_gamma(1)
+ g_gamma(2)
+ ];
+%--- rhor(std_t) = g_rhor(s_{t})
+rhortd = [
+ g_rhor(1)
+ g_rhor(1)
+ g_rhor(2)
+ g_rhor(2)
+ ];
+%--- phipi(std_t) = g_phipi(s_{t})
+phipitd = [
+ g_phipi(1)
+ g_phipi(1)
+ g_phipi(2)
+ g_phipi(2)
+ ];
+%--- phiy(std_t) = g_phiy(s_{t})
+phiytd = [
+ g_phiy(1)
+ g_phiy(1)
+ g_phiy(2)
+ g_phiy(2)
+ ];
+
+%-------------------- Filling in A(std_t)'s, B(std_t)'s, etc for an expanded gensys. -------------------%
+Astd = cell(htd,1);
+Bstd = cell(htd,1);
+Psistd = cell(htd,1);
+for si=1:htd
+ a1td = (1+g_beta*psi1td(si)*gamma0td(si));
+ a2td = psi2td(si)*((g_xi+1)/g_alpha + b/(g_lambda-b));
+ Astd{si} = [
+ -a1td a2td 0 psi2td(si) psi2td(si)*(b/(g_lambda-b)) g_beta*psi1td(si) 0
+ 0 -(g_lambda+b)/g_lambda -(g_lambda-b)/g_lambda 0 g_rhov - b/g_lambda (g_lambda-b)/g_lambda 1
+ -(1-rhortd(si))*phipitd(si) -(1-rhortd(si))*phiytd(si) 1 0 0 0 0
+ 0 0 0 1 0 0 0
+ 0 0 0 0 1 0 0
+ ];
+ Bstd{si} = [
+ -gamma1td(si) psi2td(si)*b/(g_lambda-b) 0 0 0 0 0
+ 0 -b/g_lambda 0 0 0 0 0
+ 0 0 rhortd(si) 0 0 0 0
+ 0 0 0 g_rhow 0 0 0
+ 0 0 0 0 g_rhov 0 0
+ ];
+ Psistd{si} = [
+ 0 0 0
+ 0 0 0
+ g_sigmar 0 0
+ 0 g_sigmaw 0
+ 0 0 g_sigmav
+ ];
+end
+%--- <<>>Reordering
+if (flag_swap_fps)
+ ki = 4;
+ kk = 1;
+ Atmp = Astd{ki};
+ Astd{ki} = Astd{kk};
+ Astd{kk} = Atmp;
+ Btmp = Bstd{ki};
+ Bstd{ki} = Bstd{kk};
+ Bstd{kk} = Btmp;
+ SwapM = zeros(ki,ki);
+ SwapM(kk,ki) = kk;
+ SwapM(ki,1) = 1;
+ SwapM(2,2) = 1;
+ SwapM(3,3) = 1;
+ Qtd = SwapM*Qtd*SwapM;
+end
+
+
+
+
+Aind = cell(htd,1);
+Bind = cell(htd,1);
+Psiind = cell(htd,1);
+Piind = cell(htd,1);
+G1ind = cell(htd,1);
+impactind = cell(htd,1);
+gevind = cell(htd,1);
+gevind_abs = cell(htd,1);
+euind = cell(htd,1);
+for si=1:htd
+ Aind{si} = [
+ Astd{si}
+ 1 0 0 0 0 0 0
+ 0 1 0 0 0 0 0
+ ];
+ Bind{si} = [
+ Bstd{si}
+ 0 0 0 0 0 1 0
+ 0 0 0 0 0 0 1
+ ];
+ Psiind{si} = [
+ Psistd{si}
+ 0 0 0
+ 0 0 0
+ ];
+ Piind{si} = [
+ zeros(5,2)
+ 1 0
+ 0 1
+ ];
+
+ disp('---------- Individual (possibly swapped) regimes in isolation --------------')
+ disp('Under Regime')
+ si
+
+ %[G1ind{si},Cind,impactind{si},fmat,fwt,ywt,gevind{si},euind{si}]=gensys(Aind{si}, Bind{si}, zeros(7,1), Psiind{si}, Piind{si}, divind(si));
+ %[G1ind{si},impactind{si},gevind{si},euind{si}]=msv_one(Aind{si}, Bind{si}, Psiind{si}, Piind{si});
+ [G1ind{si},impactind{si},gevind{si},euind{si}]=msv_simple(Aind{si}, Bind{si}, Psiind{si}, Piind{si});
+ [G1 impact gev eu]=msv_all(Aind{si}, Bind{si}, Psiind{si}, Piind{si});
+ %[G1ind{si},impactind{si},v2junk,gevind{si},euind{si}]=gensys_dw(Aind{si}, Bind{si}, Psiind{si}, Piind{si}, divind(si));
+
+ k=size(eu,2);
+ for i=1:k
+ gev_abs = abs(gev{i}(:,1).\gev{i}(:,2));
+ gev_abs
+ G1{i}
+ impact{i}
+ eu{i}
+ end
+end
+% disp(' ')
+% disp('----------------- Existence and uniqueness for each individual regime -----------------------')
+% for si=1:htd
+% euind{si}
+% end
+return
+
+
diff --git a/MatlabFiles/MSV/OldVersions/msv_all.m b/MatlabFiles/MSV/OldVersions/msv_all.m
new file mode 100755
index 0000000000000000000000000000000000000000..7fffafc11a21b995b35c8d85ba81484aca939a72
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/msv_all.m
@@ -0,0 +1,142 @@
+function [G1,impact,gev,eu]=msv_all(g0,g1,psi,pi)
+% function [G1,C,impact,gev,eu]=msv_all(g0,g1,psi,pi)
+% System given as
+% g0*y(t)=g1*y(t-1)+psi*z(t)+pi*eta(t),
+% with z an exogenous variable process and eta being endogenously determined
+% one-step-ahead expectational errors. Returned system is
+% y(t)=G1*y(t-1)+impact*z(t).
+% eu(1)=1 for existence, eu(2)=1 for uniqueness.
+% eu=[-2,-2] for coincident zeros.
+% Attempts all possible ordering such that all explosive roots are
+% suppressed and no complex conjugate pairs are split
+% By Daniel Waggoner -- Based on code by Christopher A. Sims
+
+realsmall=1e-10;
+
+n=size(pi,1);
+s=size(pi,2);
+ii=1;
+
+[a b q z]=qz(g0,g1);
+
+for i=1:n
+ if (abs(a(i,i)) < realsmall) & (abs(b(i,i)) < realsmall)
+ disp('Coincident zeros.')
+ eu{ii}=[-2;-2];
+ gev{ii}=[diag(a) diag(b)];
+ G1{ii}=[]
+ impact{ii}=[]
+ return
+ end
+end
+
+[a b q z]=qzsort(a,b,q,z);
+
+% determine complex conjugate pairs
+pairs=zeros(n,1);
+i=1;
+while i < n
+ if abs(b(i+1,i+1)*conj(a(i,i)) - a(i+1,i+1)*conj(b(i,i))) < realsmall
+ pairs(i)=1;
+ pairs(i+1)=-1;
+ i=i+2;
+ else
+ i=i+1;
+ end
+end
+
+% determine number unstable roots
+nunstable=0;
+i=1;
+while i <= n
+ if pairs(i) == 0
+ if (abs(b(i,i)) > abs(a(i,i)))
+ nunstable=n-i+1;
+ break;
+ else
+ i=i+1;
+ end
+ else
+ if (abs(b(i,i)) > abs(a(i,i))) | (abs(b(i+1,i+1)) > abs(a(i+1,i+1)))
+ nunstable=n-i+1;
+ break;
+ else
+ i=i+2;
+ end
+ end
+end
+
+% No bounded MSV solutions?
+if nunstable > s
+ disp('No bounded MSV solutions')
+ eu{ii}=[0;0];
+ gev{ii}=[diag(a) diag(b)];
+ G1{ii}=[]
+ impact{ii}=[]
+ return;
+end
+
+% Multiple bounded MSV solutions?
+if nunstable < s
+ disp('Multiple bounded MSV solutions should exist');
+
+ % Get roots to suppress
+ idx=zeros(n,1);
+ idx_size=zeros(n,1);
+ k=1;
+ idx(1)=n-nunstable;
+ cont=1;
+ while cont > 0
+ if (pairs(idx(k)) == -1)
+ idx_size(k)=2;
+ else
+ idx_size(k)=1;
+ end
+
+ d=nunstable;
+ for i=1:k
+ d=d+idx_size(i);
+ end
+
+ if (d == s)
+ cont=cont+1;
+ [an bn qn zn]=qzmoveindex(a,b,q,z,pairs,idx,k,n-nunstable);
+ [G1{ii},impact{ii},eu{ii}]=qzcomputesolution(an,bn,qn,zn,psi,pi,s);
+ gev{ii}=[diag(an) diag(bn)];
+ ii=ii+1;
+ end
+
+ if (d >= s)
+ if idx(k) > idx_size(k)
+ idx(k)=idx(k)-idx_size(k);
+ else
+ if k > 1
+ k=k-1;
+ idx(k)=idx(k)-idx_size(k);
+ else
+ return;
+ end
+ end
+ else
+ if idx(k) > idx_size(k)
+ idx(k+1)=idx(k)-idx_size(k);
+ k=k+1;
+ else
+ if k > 1
+ k=k-1;
+ idx(k)=idx(k)-idx_size(k);
+ else
+ return;
+ end
+ end
+ end
+ end
+end
+
+% Unique MSV solution?
+disp('Unique bounded MVS solution');
+[G1{ii},impact{ii},eu{ii}]=qzcomputesolution(a,b,q,z,psi,pi,s);
+gev{ii}=[diag(a) diag(b)];
+
+
+
diff --git a/MatlabFiles/MSV/OldVersions/msv_one.m b/MatlabFiles/MSV/OldVersions/msv_one.m
new file mode 100755
index 0000000000000000000000000000000000000000..44c79456919075887ebfc5a122a53f6ff84d6711
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/msv_one.m
@@ -0,0 +1,145 @@
+function [G1,impact,gev,eu]=msv_one(g0,g1,psi,pi)
+% function [G1,C,impact,gev,eu]=msv_one(g0,g1,psi,pi)
+% System given as
+% g0*y(t)=g1*y(t-1)+psi*z(t)+pi*eta(t),
+% with z an exogenous variable process and eta being endogenously determined
+% one-step-ahead expectational errors. Returned system is
+% y(t)=G1*y(t-1)+impact*z(t).
+% eu(1)=1 for existence, eu(2)=1 for uniqueness. eu(1)=-1 for
+% existence only with not-s.c. z; eu=[-2,-2] for coincident zeros.
+% By Daniel Waggoner -- Based on code by Christopher A. Sims
+
+realsmall=1e-10;
+
+n=size(pi,1);
+s=size(pi,2);
+
+[a b q z]=qz(g0,g1);
+
+for i=1:n
+ if (abs(a(i,i)) < realsmall) & (abs(b(i,i)) < realsmall)
+ disp('Coincident zeros.')
+ eu=[-2;-2];
+ gev=[diag(a) diag(b)];
+ G1=[]
+ impact=[]
+ return
+ end
+end
+
+[a b q z]=qzsort(a,b,q,z);
+
+% determine complex conjugate pairs
+pairs=zeros(n,1);
+i=1;
+while i < n
+ if abs(b(i+1,i+1)*conj(a(i,i)) - a(i+1,i+1)*conj(b(i,i))) < realsmall
+ pairs(i)=1;
+ pairs(i+1)=-1;
+ i=i+2;
+ else
+ i=i+1;
+ end
+end
+
+% determine number unstable roots
+nunstable=0;
+i=1;
+while i <= n
+ if pairs(i) == 0
+ if (abs(b(i,i)) > abs(a(i,i)))
+ nunstable=n-i+1;
+ break;
+ else
+ i=i+1;
+ end
+ else
+ if (abs(b(i,i)) > abs(a(i,i))) | (abs(b(i+1,i+1)) > abs(a(i+1,i+1)))
+ nunstable=n-i+1;
+ break;
+ else
+ i=i+2;
+ end
+ end
+end
+
+% No bounded MSV solutions?
+if nunstable > s
+ disp('No bounded MSV solutions')
+ eu=[0;0];
+ gev=[diag(a) diag(b)];
+ G1=[]
+ impact=[]
+ return;
+end
+
+% Multiple bounded MSV solutions?
+if nunstable < s
+ disp('Multiple bounded MSV solutions should exist');
+
+ % Get roots to suppress
+ idx=zeros(n,1);
+ idx_size=zeros(n,1);
+ k=1;
+ idx(1)=n-nunstable;
+ cont=1;
+ while cont > 0
+ if (pairs(idx(k)) == -1)
+ idx_size(k)=2;
+ else
+ idx_size(k)=1;
+ end
+
+ d=nunstable;
+ for i=1:k
+ d=d+idx_size(i);
+ end
+
+ if (d == s)
+ cont=cont+1;
+ [an bn qn zn]=qzmoveindex(a,b,q,z,pairs,idx,k,n-nunstable);
+ [G1,impact,eu]=qzcomputesolution(an,bn,qn,zn,psi,pi,s);
+ if (eu(1) == 1) & (eu(2) == 1)
+ gev=[diag(an) diag(bn)];
+ return;
+ else
+ disp('MSV solution does not exist for the ordering:');
+ abs(diag(bn)./diag(an))
+ eu
+ end
+ end
+
+ if (d >= s)
+ if idx(k) > idx_size(k)
+ idx(k)=idx(k)-idx_size(k);
+ else
+ if k > 1
+ k=k-1;
+ idx(k)=idx(k)-idx_size(k);
+ else
+ disp('No MSV solutions exist');
+ return;
+ end
+ end
+ else
+ if idx(k) > idx_size(k)
+ idx(k+1)=idx(k)-idx_size(k);
+ k=k+1;
+ else
+ if k > 1
+ k=k-1;
+ idx(k)=idx(k)-idx_size(k);
+ else
+ disp('No MSV solutions exist');
+ return;
+ end
+ end
+ end
+ end
+end
+
+% Unique MSV solution?
+disp('Unique bounded MVS solution');
+[G1,impact,eu]=qzcomputesolution(a,b,q,z,psi,pi,s);
+gev=[diag(a) diag(b)];
+
diff --git a/MatlabFiles/MSV/OldVersions/msv_simple.m b/MatlabFiles/MSV/OldVersions/msv_simple.m
new file mode 100755
index 0000000000000000000000000000000000000000..c37148fc3e6e279cf95ee4e33a669458e45cc6a4
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/msv_simple.m
@@ -0,0 +1,62 @@
+function [G1,impact,gev,eu]=msv_simple(g0,g1,psi,pi)
+% function [G1,C,impact,gev,eu]=msv_one(g0,g1,psi,pi)
+% System given as
+% g0*y(t)=g1*y(t-1)+psi*z(t)+pi*eta(t),
+% with z an exogenous variable process and eta being endogenously determined
+% one-step-ahead expectational errors. Returned system is
+% y(t)=G1*y(t-1)+impact*z(t).
+% eu(1)=1 for existence, eu(2)=1 for uniqueness.
+% eu=[-2,-2] for coincident zeros.
+% By Daniel Waggoner -- Based on code by Christopher A. Sims
+
+realsmall=1e-10;
+
+G1=[];
+impact=[];
+gev=[];
+
+n=size(pi,1);
+s=size(pi,2);
+
+[a b q z]=qz(g0,g1);
+
+for i=1:n
+ if (abs(a(i,i)) < realsmall) & (abs(b(i,i)) < realsmall)
+ disp('Coincident zeros.')
+ eu=[-2;-2];
+ gev=[diag(a) diag(b)];
+ return
+ end
+end
+
+[a b q z]=qzsort(a,b,q,z);
+
+% determine number unstable roots
+nunstable=0;
+i=n;
+while i > 0
+ if (abs(b(i,i)) > abs(a(i,i)))
+ nunstable=nunstable+1;
+ i=i-1;
+ else
+ break;
+ end
+end
+
+eu=[0;0];
+gev=[diag(a) diag(b)];
+
+% No bounded MSV solutions?
+if nunstable > s
+ disp('No bounded MSV solutions')
+ return;
+end
+
+% Multiple or unique bounded MSV solutions?
+if nunstable < s
+ disp('Multiple bounded MSV solutions should exist');
+else
+ disp('Unique bounded MVS solution');
+end
+
+[G1,impact,eu]=qzcomputesolution(a,b,q,z,psi,pi,s);
diff --git a/MatlabFiles/MSV/OldVersions/qzcomputesolution.m b/MatlabFiles/MSV/OldVersions/qzcomputesolution.m
new file mode 100755
index 0000000000000000000000000000000000000000..2312735e29e3b833ffd66d2129635022e0f5eaf9
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/qzcomputesolution.m
@@ -0,0 +1,126 @@
+function [G1,impact,eu]=qzcomputesolution(a,b,q,z,psi,pi,suppress)
+% function [G1,C,impact,eu]=qzcomputesolution(a,b,q,z,c,psi,pi,suppress)
+% System given as
+% (q'*a*z')*y(t)=(q'*b*z')*y(t-1)+psi*e(t)+pi*eta(t),
+% with e an exogenous variable process and eta an endogenously determined
+% one-step-ahead expectational error. q and z are unitary matrices and a
+% and b are upper triangular matrices that are computed via a generalized
+% Schur decomposition. It is assumed that the ordering of the system
+% is such that complex conjugate pairs are together. If a complex
+% conjugate pair is split, then G1 and impact will be complex.
+% Returned system is
+% y(t)=G1*y(t-1)+impact*z(t).
+% eu(1)=1 for existence, eu(2)=1 for uniqueness.
+%
+% The last suppress roots are suppressed.
+%
+% By Daniel Waggoner -- Based on code by Christopher A. Sims
+
+realsmall=sqrt(eps);
+
+eu=[0;0];
+G1=[];
+impact=[];
+
+n=size(pi,1);
+s=size(pi,2);
+
+q1=q(1:n-suppress,:);
+q2=q(n-suppress+1:n,:);
+
+[u1 d1 v1]=svd(q2*pi,'econ');
+if suppress > 0
+ m=sum(diag(d1) > realsmall*d1(1,1));
+else
+ m=0;
+end
+u1=u1(:,1:m);
+d1=d1(1:m,1:m);
+v1=v1(:,1:m);
+
+if (m == s)
+ eu(2)=1;
+else
+ eu(2)=0;
+end
+
+[u2 d2 v2]=svd(q2*psi,'econ');
+if suppress > 0
+ m=sum(diag(d2) > realsmall*d2(1,1));
+else
+ m=0;
+end
+u2=u2(:,1:m);
+d2=d2(1:m,1:m);
+v2=v2(:,1:m);
+
+if norm((eye(suppress) - u1*u1')*u2) < realsmall
+ eu(1)=1;
+else
+ eu(1)=0;
+ return;
+end
+
+
+% Compute impact and G0
+a11_inv=inv(a(1:n-suppress,1:n-suppress));
+x=a11_inv*q1*pi*v1*diag(1./diag(d1))*u1';
+impact=[a11_inv*q1*psi - x*q2*psi; zeros(suppress,size(psi,2))];
+
+G1=zeros(n,n);
+G1(1:n-suppress,1:n-suppress)=a11_inv*b(1:n-suppress,1:n-suppress);
+
+% this is to make the answer agree with gensys - this piece does not effect
+% the answer, at least after the initial period.
+G1(1:n-suppress,n-suppress+1:n)=a11_inv*b(1:n-suppress,n-suppress+1:n) - x*b(n-suppress+1:n,n-suppress+1:n);
+
+% Convert back to y
+impact=z*impact;
+G1=z*G1*z';
+
+% Is the solution real?
+if (suppress == 0) | (suppress == n) | (abs(b(n-suppress+1,n-suppress+1)*conj(a(n-suppress,n-suppress)) - a(n-suppress+1,n-suppress+1)*conj(b(n-suppress,n-suppress))) > realsmall)
+ impact=real(impact);
+ G1=real(G1);
+end
+
+%--------------------------------------------------------------------------
+% The code below check that a solution is obtained with stable manifold
+% z1. Should be commented out for production runs.
+%--------------------------------------------------------------------------
+% z2=z(:,n-suppress+1:n);
+% z1=null(z2');
+% A=q'*a*z';
+% B=q'*b*z';
+%
+% n=norm(A*(G1*z1) - B*z1);
+% disp('norm(A*(g1*z1) - B*z1) -- should be zero');
+% disp(n);
+%
+% eta=-v1*diag(1./diag(d1))*u1'*q2*psi;
+% n=norm(A*impact - psi - pi*eta);
+% disp('norm(A*impact - psi - pi*eta) -- should be zero');
+% disp(n);
+%
+% n=norm(z2'*G1*z1);
+% disp('norm(z2''*G1*z1) -- should be zero');
+% disp(n);
+%
+% n=norm(z2'*G1);
+% disp('norm(z2''G1) -- best if zero');
+% disp(n);
+%
+% n=norm(z2'*impact);
+% disp('norm(z2''*impact) -- should be zero');
+% disp(n);
+%
+% n=norm(G1*z2);
+% disp('norm(G1*z2)');
+% disp(n);
+%
+% disp('press any key to continue');
+% pause
+
+
+
+
diff --git a/MatlabFiles/MSV/OldVersions/qzdiv.m b/MatlabFiles/MSV/OldVersions/qzdiv.m
new file mode 100755
index 0000000000000000000000000000000000000000..b04068fe95afabf3cf4a096e90849529fef410d9
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/qzdiv.m
@@ -0,0 +1,40 @@
+function [A,B,Q,Z,v] = qzdiv(stake,A,B,Q,Z,v)
+%function [A,B,Q,Z,v] = qzdiv(stake,A,B,Q,Z,v)
+%
+% Takes U.T. matrices A, B, orthonormal matrices Q,Z, rearranges them
+% so that all cases of abs(B(i,i)/A(i,i))>stake are in lower right
+% corner, while preserving U.T. and orthonormal properties and Q'AZ' and
+% Q'BZ'. The columns of v are sorted correspondingly.
+%
+% by Christopher A. Sims
+% modified (to add v to input and output) 7/27/00
+vin = nargin==6;
+%if ~vin, v=[], end;
+if ~vin, v=[]; end;
+[n jnk] = size(A);
+root = abs([diag(A) diag(B)]);
+root(:,1) = root(:,1)-(root(:,1)<1.e-13).*(root(:,1)+root(:,2));
+root(:,2) = root(:,2)./root(:,1);
+for i = n:-1:1
+ m=0;
+ for j=i:-1:1
+ if (root(j,2) > stake | root(j,2) < -.1)
+ m=j;
+ break
+ end
+ end
+ if (m==0)
+ return
+ end
+ for k=m:1:i-1
+ [A B Q Z] = qzswitch(k,A,B,Q,Z);
+ tmp = root(k,2);
+ root(k,2) = root(k+1,2);
+ root(k+1,2) = tmp;
+ if vin
+ tmp=v(:,k);
+ v(:,k)=v(:,k+1);
+ v(:,k+1)=tmp;
+ end
+ end
+end
diff --git a/MatlabFiles/MSV/OldVersions/qzmoveindex.m b/MatlabFiles/MSV/OldVersions/qzmoveindex.m
new file mode 100755
index 0000000000000000000000000000000000000000..853b79d90263ac3d76dff67f466edab1f4f217a3
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/qzmoveindex.m
@@ -0,0 +1,17 @@
+function [a,b,q,z] = qzmoveindex(a,b,q,z,pairs,idx,k,j)
+% function [a,b,q,z] = qzmoveindex(a,b,q,z,pairs,idx,k,j)
+%
+% Takes U.T. matrices a, b, orthonormal matrices q,z, rearranges them
+% so that the indices in idx are moved up to the jth position, while
+% preserving U.T. and orthogonal properties and q'az' and q'bz'.
+%
+% by Daniel Waggoner based on code by Christopher A. Sims
+
+for i=1:k
+ [a b q z pairs]=qzslide(a,b,q,z,pairs,idx(i),j);
+ if (pairs(j) == -1)
+ j=j-2;
+ else
+ j=j-1;
+ end
+end
\ No newline at end of file
diff --git a/MatlabFiles/MSV/OldVersions/qzslide.m b/MatlabFiles/MSV/OldVersions/qzslide.m
new file mode 100755
index 0000000000000000000000000000000000000000..f9bbcb46d93fd903c5f3e3aa31c81d9ce49c2f51
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/qzslide.m
@@ -0,0 +1,80 @@
+function [a,b,q,z,pairs] = qzslide(a,b,q,z,pairs,i,j)
+% function [a,b,q,z,pairs] = qzslide(a,b,q,z,pairs,i,j)
+%
+% Takes U.T. matrices A, B, orthonormal matrices Q,Z, rearranges them
+% so that the ith diagonal element is moved to the jth position, while
+% preserving U.T. and orthogonal properties and Q'AZ' and Q'BZ'.
+%
+% The array pairs is also rearranged. Complex conjugate pairs are kept
+% adjacent. If pairs(i)=1, then position i and i+1 are complex
+% conjugate pairs. If pairs(i)=-1, then position i and i-1 are complex
+% conjugate pairs. If pairs(i)=0, then position i is real.
+%
+%
+% by Daniel Waggoner based on code by Christopher A. Sims
+
+n=size(a,1);
+
+if i == j
+ return;
+end
+
+if i < j
+ if pairs(j) == 1
+ j=j+1;
+ end
+ if pairs(i) == -1
+ j=j-1;
+ i=i-1;
+ else
+ if (pairs(i) == 1) & (j == n)
+ if (i == n-1)
+ return;
+ end
+ j=n-1;
+ end
+ end
+ while i < j
+ if (pairs(i) == 1)
+ [a b q z]=qzswitch(i+1,a,b,q,z);
+ [a b q z]=qzswitch(i,a,b,q,z);
+ pairs(i)=pairs(i+2);
+ pairs(i+1)=1;
+ pairs(i+2)=-1;
+ else
+ [a b q z]=qzswitch(i,a,b,q,z);
+ pairs(i)=pairs(i+1);
+ pairs(i+1)=0;
+ end
+ i=i+1;
+ end
+else
+ if pairs(j) == -1
+ j=j-1;
+ end
+ if pairs(i) == 1
+ j=j+1;
+ i=i+1;
+ else
+ if (pairs(i) == -1) & (j == 1)
+ if (i == 2)
+ return;
+ end
+ j=2;
+ end
+ end
+ while i > j
+ if (pairs(i) == -1)
+ [a b q z]=qzswitch(i-2,a,b,q,z);
+ [a b q z]=qzswitch(i-1,a,b,q,z);
+ pairs(i)=pairs(i-2);
+ pairs(i-2)=1;
+ pairs(i-1)=-1;
+ else
+ [a b q z]=qzswitch(i-1,a,b,q,z);
+ pairs(i)=pairs(i-1);
+ pairs(i-1)=0;
+ end
+ i=i-1;
+ end
+end
diff --git a/MatlabFiles/MSV/OldVersions/qzsort.m b/MatlabFiles/MSV/OldVersions/qzsort.m
new file mode 100755
index 0000000000000000000000000000000000000000..7f24331a9650d14c0ae42c25fb2f1a67bed4c754
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/qzsort.m
@@ -0,0 +1,21 @@
+function [A,B,Q,Z] = qzsort(A,B,Q,Z)
+%function [A,B,Q,Z] = qzsort(stake,A,B,Q,Z)
+%
+% Takes U.T. matrices A, B, orthonormal matrices Q,Z, rearranges them
+% so that abs(B(i,i)/A(i,i)) are increasing, while preserving U.T. and
+% orthonormal properties and Q'AZ' and % Q'BZ'.
+%
+% by Daniel Waggoner based on code by Christopher A. Sims
+
+n=size(A,1);
+
+for i=2:n
+ for j=i:-1:2
+ if abs(A(j,j)*B(j-1,j-1)) > abs(A(j-1,j-1)*B(j,j))
+ [A B Q Z]=qzswitch(j-1,A,B,Q,Z);
+ else
+ break;
+ end
+ end
+end
+
diff --git a/MatlabFiles/MSV/OldVersions/qzswitch.m b/MatlabFiles/MSV/OldVersions/qzswitch.m
new file mode 100755
index 0000000000000000000000000000000000000000..93c36640cc3488a66dc6c5261b8778e45c314c94
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/qzswitch.m
@@ -0,0 +1,60 @@
+function [A,B,Q,Z] = qzswitch(i,A,B,Q,Z)
+%function [A,B,Q,Z] = qzswitch(i,A,B,Q,Z)
+%
+% Takes U.T. matrices A, B, orthonormal matrices Q,Z, interchanges
+% diagonal elements i and i+1 of both A and B, while maintaining
+% Q'AZ' and Q'BZ' unchanged. If diagonal elements of A and B
+% are zero at matching positions, the returned A will have zeros at both
+% positions on the diagonal. This is natural behavior if this routine is used
+% to drive all zeros on the diagonal of A to the lower right, but in this case
+% the qz transformation is not unique and it is not possible simply to switch
+% the positions of the diagonal elements of both A and B.
+ realsmall=sqrt(eps)*10;
+%realsmall=1e-3;
+a = A(i,i); d = B(i,i); b = A(i,i+1); e = B(i,i+1);
+c = A(i+1,i+1); f = B(i+1,i+1);
+ % A(i:i+1,i:i+1)=[a b; 0 c];
+ % B(i:i+1,i:i+1)=[d e; 0 f];
+if (abs(c)<realsmall & abs(f)<realsmall)
+ if abs(a)<realsmall
+ % l.r. coincident 0's with u.l. of A=0; do nothing
+ return
+ else
+ % l.r. coincident zeros; put 0 in u.l. of a
+ wz=[b; -a];
+ wz=wz/sqrt(wz'*wz);
+ wz=[wz [wz(2)';-wz(1)'] ];
+ xy=eye(2);
+ end
+elseif (abs(a)<realsmall & abs(d)<realsmall)
+ if abs(c)<realsmall
+ % u.l. coincident zeros with l.r. of A=0; do nothing
+ return
+ else
+ % u.l. coincident zeros; put 0 in l.r. of A
+ wz=eye(2);
+ xy=[c -b];
+ xy=xy/sqrt(xy*xy');
+ xy=[[xy(2)' -xy(1)'];xy];
+ end
+else
+ % usual case
+ wz = [c*e-f*b, (c*d-f*a)'];
+ xy = [(b*d-e*a)', (c*d-f*a)'];
+ n = sqrt(wz*wz');
+ m = sqrt(xy*xy');
+ if m<eps*100
+ % all elements of A and B proportional
+ return
+ end
+ wz = n\wz;
+ xy = m\xy;
+ wz = [wz; -wz(2)', wz(1)'];
+ xy = [xy;-xy(2)', xy(1)'];
+end
+A(i:i+1,:) = xy*A(i:i+1,:);
+B(i:i+1,:) = xy*B(i:i+1,:);
+A(:,i:i+1) = A(:,i:i+1)*wz;
+B(:,i:i+1) = B(:,i:i+1)*wz;
+Z(:,i:i+1) = Z(:,i:i+1)*wz;
+Q(i:i+1,:) = xy*Q(i:i+1,:);
\ No newline at end of file
diff --git a/MatlabFiles/MSV/OldVersions/test_gensys.m b/MatlabFiles/MSV/OldVersions/test_gensys.m
new file mode 100755
index 0000000000000000000000000000000000000000..f8ae311cd2304f84cb4bedfc81871f8a161a837d
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/test_gensys.m
@@ -0,0 +1,53 @@
+r=2;
+s=2;
+n=6;
+explosive=s;
+
+div=1.0;
+
+cont=1;
+while cont > 0
+ [q tmp]=qr(rand(n,n));
+ [z tmp]=qr(rand(n,n));
+ a=q*diag(ones(n,1)+rand(n,1))*z;
+ %a=q*z;
+ d=rand(n,1);
+ d(1:explosive)=d(1:explosive)+ones(explosive,1);
+ b=q*diag(d)*z;
+
+ psi=rand(n,r);
+ pi=rand(n,s);
+ c=zeros(n,1);
+
+ [G1,C,impact,fmat,fwt,ywt,gev,eu]=gensys(a,b,c,psi,pi,div);
+ [dwG1,dwimpact,dwgev,dweu]=dw_gensys(a,b,psi,pi,div);
+
+ G1
+ dwG1
+
+ impact
+ dwimpact
+
+ eu
+ dweu
+
+ if eu(1) == 1
+ cont=0;
+ end
+end
+
+% Simulate
+y0=rand(n,1)
+n_sim=10;
+for i=1:10
+ epsilon=rand(r,1);
+ y=G1*y0+impact*epsilon;
+ dwy=dwG1*y0+dwimpact*epsilon;
+
+ y
+ dwy
+ norm(y-dwy)
+ pause
+
+ y0=dwy;
+end
\ No newline at end of file
diff --git a/MatlabFiles/MSV/OldVersions/test_msv.m b/MatlabFiles/MSV/OldVersions/test_msv.m
new file mode 100755
index 0000000000000000000000000000000000000000..c408620adca36353d8b6b98e5b78fcf96b1fc4b5
--- /dev/null
+++ b/MatlabFiles/MSV/OldVersions/test_msv.m
@@ -0,0 +1,35 @@
+r=2;
+s=2;
+n=6;
+explosive=s;
+
+div=1.0;
+
+cont=1;
+while cont > 0
+ [q tmp]=qr(rand(n,n));
+ [z tmp]=qr(rand(n,n));
+ a=q*diag(ones(n,1)+rand(n,1))*z;
+ %a=q*z;
+ d=rand(n,1);
+ d(1:explosive)=d(1:explosive)+ones(explosive,1);
+ b=q*diag(d)*z;
+
+ psi=rand(n,r);
+ pi=rand(n,s);
+
+ [G1,G2,gev,eu]=msv_all(a,b,psi,pi);
+
+ k=size(eu,2);
+
+ for i=1:k
+ eu{i}
+ abs(gev{i}(:,2)./gev{i}(:,1))
+ G1{i}
+ G2{i}
+ i
+ k
+ pause
+ end
+end
+
diff --git a/MatlabFiles/MSV/fwz_msv_msre.m b/MatlabFiles/MSV/fwz_msv_msre.m
new file mode 100755
index 0000000000000000000000000000000000000000..5b97593340fcbd8b010922b4b22947f4172d3817
--- /dev/null
+++ b/MatlabFiles/MSV/fwz_msv_msre.m
@@ -0,0 +1,125 @@
+function [F1, F2, G1, G2, V, err] = fwz_msv_msre(P, A, B, Psi, s, x, max_count, tol)
+%[ F1 F2 G1 G2 V err ] = fwz_msv_msre(P, A, B, Psi, s, x, max_count, tol)
+% Computes MSV solution of
+%
+% A(s(t) x(t) = B(s(t) x(t-1) + Psi(s(t)) epsilon(t) + Pi eta(t)
+%
+% using Newton's Method. Assumes that Pi' = [zeros(s,n-s) eye(s)] and
+% that A{i} is invertible. P is the transition matrix and P(i,j) is the
+% probability that s(t+1)=j given that s(t)=i. Note that the rows of P
+% must sum to one. x is the initial value and if not passed is set to
+% zero. max_count is the maximum number of iterations on Newton's method
+% before failing and tol is the convergence criterion.
+%
+% The solution is of the form
+%
+% x(t) = V{s(t)}*F1{s(t)}*x(t-1) + V{s(t)}*G1{s(t)}*epsilon(t)
+%
+% eta(t) = F2{s(t)}*x(t-1) + G2{s(t)}*epsilont(t)
+%
+% A positive value of err is the number of iterations needed to obtain
+% convergence and indicates success. A negitive value of err is the number of
+% iterations before the method terminated without convergence and indicates
+% failure.
+%
+
+h=size(P,1);
+n=size(A{1},1);
+r=size(Psi{1},2);
+
+if (nargin <= 7) || (tol <= 0)
+ tol=1e-5;
+end
+
+if (nargin <= 6) || (max_count <= 0)
+ max_count=1000;
+end
+
+if nargin <= 5
+ x=zeros(h*s*(n-s),1);
+end
+f=zeros(h*s*(n-s),1);
+
+Is=eye(s);
+I=eye(n-s);
+
+U=cell(h,1);
+for i=1:h
+ U{i}=inv(A{i});
+end
+C=cell(h,h);
+for i=1:h
+ for j=1:h
+ C{i,j}=P(i,j)*B{j}*U{i};
+ end
+end
+
+cont=true;
+count=1;
+D=zeros(h*s*(n-s),h*s*(n-s));
+X=cell(h,1);
+for i=1:h
+ X{i}=reshape(x((i-1)*s*(n-s)+1:i*s*(n-s)),s,n-s);
+end
+while cont
+ for i=1:h
+ for j=1:h
+ W1=C{i,j}*[I; -X{i}];
+ W2=W1(1:n-s,:);
+ D((i-1)*s*(n-s)+1:i*s*(n-s),(j-1)*s*(n-s)+1:j*s*(n-s)) = kron(W2',Is);
+ if i == j
+ W1=zeros(s,n);
+ for k=1:h
+ W1=W1+[X{k} Is]*C{i,k};
+ end
+ W2=-W1(:,n-s+1:end);
+ D((i-1)*s*(n-s)+1:i*s*(n-s),(j-1)*s*(n-s)+1:j*s*(n-s)) = D((i-1)*s*(n-s)+1:i*s*(n-s),(j-1)*s*(n-s)+1:j*s*(n-s)) + kron(I,W2);
+ end
+ end
+ end
+
+ for i=1:h
+ mf=zeros(s,n-s);
+ for j=1:h
+ mf=mf+[X{j} Is]*C{i,j}*[I; -X{i}];
+ end
+ f((i-1)*s*(n-s)+1:i*s*(n-s))=reshape(mf,s*(n-s),1);
+ end
+
+ y=D\f;
+ x=x - y;
+
+ if (count > max_count) || (norm(f) < tol)
+ cont=false;
+ else
+ count=count+1;
+ for i=1:h
+ X{i}=reshape(x((i-1)*s*(n-s)+1:i*s*(n-s)),s,n-s);
+ end
+ end
+end
+
+if (norm(f) < tol)
+ err=count;
+else
+ err=-count;
+end
+
+F1=cell(h,1);
+F2=cell(h,1);
+G1=cell(h,1);
+G2=cell(h,1);
+V=cell(h,1);
+pi=[zeros(n-s,s); Is];
+for i=1:h
+ X=reshape(x((i-1)*s*(n-s)+1:i*s*(n-s)),s,n-s);
+ V{i}=U{i}*[I; -X];
+ W=[A{i}*V{i} pi];
+ F=W\B{i};
+ F1{i}=F(1:n-s,:);
+ F2{i}=F(n-s+1:end,:);
+ G=W\Psi{i};
+ G1{i}=G(1:n-s,:);
+ G2{i}=G(n-s+1:end,:);
+end
+
\ No newline at end of file
diff --git a/MatlabFiles/MSV/fwz_pi2eye.m b/MatlabFiles/MSV/fwz_pi2eye.m
new file mode 100644
index 0000000000000000000000000000000000000000..eb58610077f5080b39ea16f8ed8f19352b3fff68
--- /dev/null
+++ b/MatlabFiles/MSV/fwz_pi2eye.m
@@ -0,0 +1,20 @@
+function [A_new, B_new, Psi_new] = fwz_pi2eye(A, B, Psi, Pi)
+% [A, B, Psi] = fwz_pi2eye(A, B, Psi, Pi)
+% Converts the representation
+%
+% A(s(t) x(t) = B(s(t) x(t-1) + Psi(s(t)) epsilon(t) + Pi{s(t)} eta(t)
+%
+% to a form compatible with fwz_msv_msre(). This is the form
+%
+% A(s(t) x(t) = B(s(t) x(t-1) + Psi(s(t)) epsilon(t) + Pi eta(t)
+%
+% where Pi' = [zeros(s,n-s) eye(s)]
+%
+
+for i=1:h
+ [Q,R] = qr(Pi{i});
+ W=[zeros(n-s,s) I; inv(R(1:s,1:s)); zeros(s,n-s)]*Q';
+ A_new{i}=W*A{i};
+ B_new{i}=W*B{i};
+ Psi_new{i}=W*Psi{i};
+end
\ No newline at end of file
diff --git a/MatlabFiles/MSV/msv_all_complex_AR.m b/MatlabFiles/MSV/msv_all_complex_AR.m
new file mode 100755
index 0000000000000000000000000000000000000000..3ad0abddcc6caed9624bbea5d2b41fce42827d6d
--- /dev/null
+++ b/MatlabFiles/MSV/msv_all_complex_AR.m
@@ -0,0 +1,112 @@
+function [G1,Impact,gev,z2,err]=msv_all_complex_AR(A,B,Psi,Pi)
+% System given as
+%
+% A*y(t) = B*y(t-1) + psi*epsilon(t) + pi*eta(t),
+%
+% with epsilon(t) an exogenous process and eta(t) endogenously determined
+% one-step-ahead expectational errors. Returned solutions are
+%
+% y(t)=G1{i}*y(t-1)+Impact{i}*epsilon(t).
+%
+% The span of the each of the solutions returned will be n-s and the
+% solution will be uniquely determined by its span. gev returns the
+% generalized eigenvalues, of which the last s were suppressed. z2 is the
+% subspace that is perpendicular to the span of the solution.
+%
+% err is the error return.
+% err = 0 -- success, at least one msv solution found.
+% err = 1 -- no msv solutions
+% err = 2 -- degenerate system
+%
+% Solutions are obtained via calls to qzcomputesolution().
+%
+% By Daniel Waggoner
+
+realsmall=sqrt(eps);
+
+n=size(Pi,1);
+s=size(Pi,2);
+
+[a,b,q,z]=qz(A,B,'complex');
+
+G1=cell(0,1);
+Impact=cell(1,1);
+gev=cell(0,1);
+gev{1}=[diag(a) diag(b)];
+z2=cell(0,1);
+err=1;
+
+% determine if the system is degenerate and count the number of diagonal
+% elements of a that are non-zero.
+kk=0;
+for i=1:n
+ if abs(a(i,i)) < realsmall
+ if abs(b(i,i)) < realsmall
+ disp('Coincident zeros.')
+ err=2;
+ return
+ end
+ else
+ kk=kk+1;
+ end
+end
+
+% too many roots must be suppressed. no msv type solutions.
+if (n-kk > s)
+ return
+end
+
+% sort the generalized eigenvalues
+[d,id]=sort(abs(diag(b)./diag(a)),'descend');
+[id,idx]=sort(id);
+[a0,b0,q0,z0]=ordqz(a,b,q,z,idx);
+a=a0; b=b0; q=q0; z=z0;
+
+idx=ones(n,1);
+for i=0:s-1
+ idx(n-i)=0;
+end
+
+ii=0;
+cont=1;
+cntmax = 500;
+cntnumber = 1;
+while (cont == 1) && (cntnumber <=cntmax)
+ % compute solution
+ [g1,impact,eu]=qzcomputesolution(a,b,q,z,Psi,Pi,s);
+
+ % save solution if unique
+ if (eu == [1;1])
+ ii=ii+1;
+ G1{ii,1}=g1;
+ Impact{ii,1}=impact;
+ gev{ii,1}=[diag(a) diag(b)];
+ z2{ii,1}=z(:,n-s+1:n);
+ err=0;
+ end
+
+ % increment idx
+ cont=0;
+ jj=1;
+ while (jj < kk) & (idx(jj) == 0)
+ idx(jj)=1;
+ jj=jj+1;
+ end
+ for j=jj+1:kk
+ if idx(j) == 0
+ idx(j)=1;
+ cont=1;
+ break;
+ end
+ end
+ for i=1:jj
+ idx(j-i)=0;
+ end
+
+ % reorder roots
+ [a,b,q,z]=ordqz(a0,b0,q0,z0,idx);
+
+ %--- Ad hoc addtion.
+ cntnumber = cntnumber+1;
+end
+
diff --git a/MatlabFiles/MSV/readme_msv.prn b/MatlabFiles/MSV/readme_msv.prn
new file mode 100755
index 0000000000000000000000000000000000000000..7b8ece011a5ccae70f64ed072ea391c8dbc086dc
--- /dev/null
+++ b/MatlabFiles/MSV/readme_msv.prn
@@ -0,0 +1,35 @@
+**********************************************************************
+************************** Important Notes ***************************
+**********************************************************************
+When eu=[0,1] in the original gensys setup, it means that in our MSV setting that exogenous persistence is put
+ in a wrong place when we reorder the roots. Thus, there is no such [0,1] return in our MSV solution.
+
+When eu=[1,0] in the original gensys setup, it means that the MSV solution must exist with another order. See
+ example3_10_sw.m in D:\ZhaData\WorkDisk\LiuWZ\ZhaNotes\UnderstandingMSV\OneDimentionCaseImportant.
+
+With our MSV solution, we abuse the notation eu=[1,0], which really means that we have multiple MSV solutions.
+ But multiple MSV solutions can occur only in complex numbers. The imaginary part of the solution can be scaled
+ arbitrarily, which gives sunspot solutions. Thus, the real MSV solution will be unique.
+
+If we encourter eu=[3,?], it means that all MSV solutions are explosive. Thus, no stable MSV solution.
+
+
+
+========================= New files for MSV solutions ==============================
+msv_complex_AR.m: gives an MSV solution (upon the order that first works) with a flag indicating if [1,0]. If
+ [1;0], the complex solution exits and thus the MSV solution is NOT unique.
+ AR: autoregressive or gensys form;
+msv_all_complex_AR.m: gives all MSV-like solutions (including complex solutions). That is, all the solutions that
+ come from the undetermined coefficients method.
+
+
+========================= Old files for MSV solutions ==============================
+msv_simple.m: works only for the first order if an msv solution (even if it is non-unique where the solution is
+ complex) exist. It does not try the second order (where the exogenous persistence may be larger
+ than the endogenous persistence).
+
+msv_one.m: works only for the unqiue MSV solution (thus, no complex solution) for the order that comes first. If no
+ solution is found for all orders, then no unique MSV solution (but there may be multiple MSV solutions
+ such as a complex solution).
+
+msv_all.m: gives all MSV-like real solutions (thus, some solutions are NOT an MSV solution).
diff --git a/MatlabFiles/MSV/verify_solution.m b/MatlabFiles/MSV/verify_solution.m
new file mode 100644
index 0000000000000000000000000000000000000000..dc0737f24cd88bcdcbeeda0406e6386423308f4c
--- /dev/null
+++ b/MatlabFiles/MSV/verify_solution.m
@@ -0,0 +1,52 @@
+function diff = verify_solution(P,A,B,Psi,s,F1,F2,G1,G2,V)
+%
+% Verifies that
+%
+% x(t) = V(s(t))(F1(s(t))x(t-1) + G1(s(t))epsilon(t)
+% eta(t) = F2(s(t))x(t-1) + G2(s(t))epsilon(t)
+%
+% is a solution of
+%
+% A(s(t)) x(t) = B(s(t)) x(t-1) + Psi(s(t) epsilon(t) + Pi eta(t)
+%
+% where Pi' = [zeros(s,n-s) eye(s)] by verifying that
+%
+% [A(j)*V(j) Pi] [ F1(j)
+% F2(j) ] = B(j)
+%
+% [A(j)*V(j) Pi] [ G1(j)
+% G2(j) ] = Psi(j)
+%
+% (P(i,1)*F2(1) + ... + P(i,h)*F2(h))*V(i) = 0
+%
+
+h=size(P,1);
+n=size(A{1},1);
+r=size(Psi{1},2);
+
+diff=0;
+Pi=[zeros(n-s,s); eye(s)];
+for j=1:h
+ X=inv([A{j}*V{j} Pi]);
+
+ tmp=norm(X*B{j} - [F1{j}; F2{j}]);
+ if tmp > diff
+ diff=tmp;
+ end
+
+ tmp=norm(X*Psi{j} - [G1{j}; G2{j}]);
+ if (tmp > diff)
+ diff=tmp;
+ end
+end
+
+for i=1:h
+ X=zeros(s,n);
+ for j=1:h
+ X=X+P(i,j)*F2{j};
+ end
+ tmp=norm(X*V{i});
+ if tmp > diff
+ diff=tmp;
+ end;
+end
diff --git a/MatlabFiles/fn_multigraphn_3g_shadedbands.m b/MatlabFiles/fn_multigraphn_3g_shadedbands.m
new file mode 100755
index 0000000000000000000000000000000000000000..4a68388d53c4d78eaf15d38277e3b328a3a86d5d
--- /dev/null
+++ b/MatlabFiles/fn_multigraphn_3g_shadedbands.m
@@ -0,0 +1,171 @@
+function scaleout = fn_multigraphn_3g_shadedbands(imfn,...
+ xlab,ylab,tstring,XTick,YTickIndx,scaleIndx,nrowg,ncolg)
+%Searching <<>> for ad hoc and specific changes.
+%
+%Stacking n sets of impulse responses in one graph. See fn_multigraph2_ver2.m.
+% imfn: row: "nstp" time horizon (in the graphics),
+% column: "nrow "variables such as responses (row in the graphics),
+% 3rd D: across "ncol" different situations such as shocks (column in the graphics),
+% 4th D: low band, estimate, high band (in each graph).
+% xlab: x-axis labels on the top
+% ylab: y-axis labels on the left
+% tstring: string for time (e.g., month or quarter) -- x-axis labels at the bottom
+% YTickIndx: 1: enable YTick; 0: disable;
+% To get a better picture, it is sometimes better to set YtickIndx to zero.
+% scaleIndx: 1: enable scale along Y-axis; 0: disable
+% nrowg: number of rows in the graph
+% ncolg: number of columns in the graph
+%
+% See imrgraph, imcerrgraph, imrerrgraph
+nstp = size(imfn,1);
+nrow = size(imfn,2);
+ncol = size(imfn,3);
+nmodels = size(imfn,4);
+t = 1:nstp;
+treverse = fliplr(t);
+
+if (nargin < 9)
+ nrowg = nrow;
+ ncolg = ncol;
+end
+if (nrowg*ncolg < nrow*ncol)
+ nrowg
+ ncolg
+ nrow
+ ncol
+
+ error('fn_multigraphn_3g_shadedbands.m: nrowg*ncolg must be greater than nrow*ncol')
+end
+
+
+tempmax=zeros(ncol,1);
+tempmin=zeros(ncol,1);
+maxval=zeros(nrow,1);
+minval=zeros(nrow,1);
+for i = 1:nrow
+ for j = 1:ncol
+ tempmax(j) = -realmax;
+ tempmin(j) = realmax;
+ for k=1:nmodels
+ jnk = max(imfn(:,i,j,k));
+ tempmax(j) = max([jnk tempmax(j)]);
+ %
+ jnk = min(imfn(:,i,j,k));
+ tempmin(j) = min([jnk tempmin(j)]);
+ end
+ end
+ maxval(i)=max(tempmax);
+ minval(i)=min(tempmin);
+end
+
+scaleout = [maxval(:) minval(:)];
+
+%--------------
+% Column j: Shock 1 to N; Row i: Responses to
+%-------------
+%figure
+
+
+rowlabel = 1;
+for i = 1:nrow % column: from top to bottom
+ columnlabel = 1;
+
+ if minval(i)<0
+ if maxval(i)<=0
+ yt=[minval(i) 0];
+ else
+ yt=[minval(i) 0 maxval(i)];
+ end
+ else % (minval(i) >=0)
+ if maxval(i) > 0
+ yt=[0 maxval(i)];
+ else % (identically zero responses)
+ yt=[-1 0 1];
+ end
+ end
+
+
+ scale=[1 nstp minval(i) maxval(i)];
+
+ for j = 1:ncol % row: from left to right
+ k1=(i-1)*ncol+j;
+ subplot(nrowg,ncolg,k1)
+
+ %set(0,'DefaultAxesColorOrder',[1 0 0;0 1 0;0 0 1],...
+ % 'DefaultAxesLineStyleOrder','-|--|:')
+ %set(0,'DefaultAxesLineStyleOrder','-|--|:|-.')
+ %---<<>>
+ %set(0,'DefaultAxesColorOrder',[0 0 0],...
+ % 'DefaultAxesLineStyleOrder','-.|-.|-|--|-.|-*|--o|:d')
+
+ nseries = zeros(nstp, nmodels);
+ for k=1:nmodels
+ nseries(:,k) = imfn(:,i,j,k);
+ end
+
+ %---<<>>
+ fill([t treverse],[nseries(:,[3])' fliplr(nseries(:,[1])')],[0.8,0.8,0.8],'EdgeColor','none');
+ %plot(t,nseries(:,[1 3]), '-.k','LineWidth',1.0); %,'Color','k');
+ hold on
+ plot(t,nseries(:,[2]), '-k','LineWidth',1.5);
+ %plot(t,nseries(:,[4]), '--k','LineWidth',1.6);
+ hold off
+
+ %set(gca,'LineStyleOrder','-|--|:|-.')
+ %set(gca,'LineStyleOrder',{'-*',':','o'})
+ grid;
+ if scaleIndx
+ axis(scale);
+ end
+ %set(gca,'YLim',[minval(i) maxval(i)])
+ %
+ set(gca,'XTick',XTick)
+ if YTickIndx
+ set(gca,'YTick',yt)
+ end
+
+ if i<nrow
+ set(gca,'XTickLabelMode','manual','XTickLabel',[])
+ end
+ %set(gca,'XTickLabel',' ');
+ if (scaleIndx) && (j>1)
+ set(gca,'YTickLabel',' ');
+ end
+ if rowlabel == 1
+ %title(['x' num2str(j)])
+ %title(eval(['x' num2str(j)]))
+ if (~isempty(xlab)), title(char(xlab(j))), end
+ end
+ if columnlabel == 1
+ %ylabel(['x' num2str(i)])
+ %ylabel(eval(['x' num2str(i)]))
+ if (~isempty(ylab)), ylabel(char(ylab(i))), end
+ end
+ if (i==nrow) && (~isempty(tstring))
+ xlabel(tstring)
+ end
+ columnlabel = 0;
+ end
+ rowlabel = 0;
+end
+
+
+%Order of line styles and markers used in a plot.
+%This property specifies which line styles and markers to use and in what order
+%when creating multiple-line plots. For example,set(gca,'LineStyleOrder', '-*|:|o')sets LineStyleOrder to solid line with asterisk
+%marker, dotted line, and hollow circle marker. The default is (-), which specifies
+%a solid line for all data plotted. Alternatively, you can create a cell array
+%of character strings to define the line styles:set(gca,'LineStyleOrder',{'-*',':','o'})MATLAB supports four line styles, which you can specify any number of
+%times in any order. MATLAB cycles through the line styles only after using
+%all colors defined by the ColorOrder property. For example,
+%the first eight lines plotted use the different colors defined by ColorOrder with
+%the first line style. MATLAB then cycles through the colors again, using the
+%second line style specified, and so on.You can also specify line style and color directly with the plot and plot3 functions
+%or by altering the properties of theline or
+%lineseries objects after creating the graph. High-Level Functions and LineStyleOrderNote that, if the axes NextPlot property is set
+%to replace (the default), high-level functions like plot reset
+%the LineStyleOrder property before determining the line
+%style to use. If you want MATLAB to use a LineStyleOrder that
+%is different from the default, set NextPlot to replacechildren. Specifying a Default LineStyleOrderYou can also specify your own default LineStyleOrder.
+%For example, this statementset(0,'DefaultAxesLineStyleOrder',{'-*',':','o'})
+%creates a default value for
diff --git a/MatlabFiles/pathdef.m b/MatlabFiles/pathdef.m
deleted file mode 100644
index 7dd3196161fd10a4efdd00f5cd17e1c407edf90b..0000000000000000000000000000000000000000
--- a/MatlabFiles/pathdef.m
+++ /dev/null
@@ -1,247 +0,0 @@
-function p = pathdef
-%PATHDEF Search path defaults.
-% PATHDEF returns a string that can be used as input to MATLABPATH
-% in order to set the path.
-
-
-% Copyright 1984-2007 The MathWorks, Inc.
-% $Revision: 1.4.2.2 $ $Date: 2007/06/07 14:45:14 $
-
-
-% DO NOT MODIFY THIS FILE. IT IS AN AUTOGENERATED FILE.
-% EDITING MAY CAUSE THE FILE TO BECOME UNREADABLE TO
-% THE PATHTOOL AND THE INSTALLER.
-
-p = [...
-%%% BEGIN ENTRIES %%%
- '/Users/tzha/ZhaData/TZCcode/cstz:', ...
- matlabroot,'/toolbox/matlab/general:', ...
- matlabroot,'/toolbox/matlab/ops:', ...
- matlabroot,'/toolbox/matlab/lang:', ...
- matlabroot,'/toolbox/matlab/elmat:', ...
- matlabroot,'/toolbox/matlab/randfun:', ...
- matlabroot,'/toolbox/matlab/elfun:', ...
- matlabroot,'/toolbox/matlab/specfun:', ...
- matlabroot,'/toolbox/matlab/matfun:', ...
- matlabroot,'/toolbox/matlab/datafun:', ...
- matlabroot,'/toolbox/matlab/polyfun:', ...
- matlabroot,'/toolbox/matlab/funfun:', ...
- matlabroot,'/toolbox/matlab/sparfun:', ...
- matlabroot,'/toolbox/matlab/scribe:', ...
- matlabroot,'/toolbox/matlab/graph2d:', ...
- matlabroot,'/toolbox/matlab/graph3d:', ...
- matlabroot,'/toolbox/matlab/specgraph:', ...
- matlabroot,'/toolbox/matlab/graphics:', ...
- matlabroot,'/toolbox/matlab/uitools:', ...
- matlabroot,'/toolbox/matlab/strfun:', ...
- matlabroot,'/toolbox/matlab/imagesci:', ...
- matlabroot,'/toolbox/matlab/iofun:', ...
- matlabroot,'/toolbox/matlab/audiovideo:', ...
- matlabroot,'/toolbox/matlab/timefun:', ...
- matlabroot,'/toolbox/matlab/datatypes:', ...
- matlabroot,'/toolbox/matlab/verctrl:', ...
- matlabroot,'/toolbox/matlab/codetools:', ...
- matlabroot,'/toolbox/matlab/helptools:', ...
- matlabroot,'/toolbox/matlab/demos:', ...
- matlabroot,'/toolbox/matlab/timeseries:', ...
- matlabroot,'/toolbox/matlab/hds:', ...
- matlabroot,'/toolbox/matlab/guide:', ...
- matlabroot,'/toolbox/matlab/plottools:', ...
- matlabroot,'/toolbox/local:', ...
- matlabroot,'/toolbox/shared/controllib:', ...
- matlabroot,'/toolbox/shared/dastudio:', ...
- matlabroot,'/toolbox/matlab/datamanager:', ...
- matlabroot,'/toolbox/simulink/simulink:', ...
- matlabroot,'/toolbox/simulink/simulink/slresolve:', ...
- matlabroot,'/toolbox/simulink/blocks:', ...
- matlabroot,'/toolbox/simulink/components:', ...
- matlabroot,'/toolbox/simulink/fixedandfloat:', ...
- matlabroot,'/toolbox/simulink/fixedandfloat/fxpdemos:', ...
- matlabroot,'/toolbox/simulink/fixedandfloat/obsolete:', ...
- matlabroot,'/toolbox/simulink/simdemos:', ...
- matlabroot,'/toolbox/simulink/simdemos/aerospace:', ...
- matlabroot,'/toolbox/simulink/simdemos/automotive:', ...
- matlabroot,'/toolbox/simulink/simdemos/simfeatures:', ...
- matlabroot,'/toolbox/simulink/simdemos/simgeneral:', ...
- matlabroot,'/toolbox/simulink/dee:', ...
- matlabroot,'/toolbox/shared/dastudio/depviewer:', ...
- matlabroot,'/toolbox/stateflow/stateflow:', ...
- matlabroot,'/toolbox/rtw/rtw:', ...
- matlabroot,'/toolbox/shared/sigbldr:', ...
- matlabroot,'/toolbox/simulink/simulink/modeladvisor:', ...
- matlabroot,'/toolbox/simulink/simulink/modeladvisor/fixpt:', ...
- matlabroot,'/toolbox/simulink/simulink/MPlayIO:', ...
- matlabroot,'/toolbox/simulink/simulink/dataobjectwizard:', ...
- matlabroot,'/toolbox/shared/hdlshared:', ...
- matlabroot,'/toolbox/rtw/accel:', ...
- matlabroot,'/toolbox/rtw/rtwdemos:', ...
- matlabroot,'/toolbox/rtw/rtwdemos/rsimdemos:', ...
- matlabroot,'/toolbox/rtw/targets/asap2/asap2:', ...
- matlabroot,'/toolbox/rtw/targets/asap2/asap2/user:', ...
- matlabroot,'/toolbox/rtw/targets/common/can/blocks:', ...
- matlabroot,'/toolbox/rtw/targets/common/configuration/resource:', ...
- matlabroot,'/toolbox/rtw/targets/common/tgtcommon:', ...
- matlabroot,'/toolbox/rtw/targets/connectivity:', ...
- matlabroot,'/toolbox/rtw/targets/pil:', ...
- matlabroot,'/toolbox/rtw/rtw/datadiff/GUI:', ...
- matlabroot,'/toolbox/rtw/rtw/datadiff/GUI/Icons:', ...
- matlabroot,'/toolbox/rtw/rtw/datadiff/API:', ...
- matlabroot,'/toolbox/rtw/rtw/cgv/API:', ...
- matlabroot,'/toolbox/stateflow/sfdemos:', ...
- matlabroot,'/toolbox/stateflow/coder:', ...
- matlabroot,'/toolbox/bioinfo/bioinfo:', ...
- matlabroot,'/toolbox/bioinfo/biolearning:', ...
- matlabroot,'/toolbox/bioinfo/microarray:', ...
- matlabroot,'/toolbox/bioinfo/mass_spec:', ...
- matlabroot,'/toolbox/bioinfo/proteins:', ...
- matlabroot,'/toolbox/bioinfo/biomatrices:', ...
- matlabroot,'/toolbox/bioinfo/biodemos:', ...
- matlabroot,'/toolbox/bioinfo/graphtheory:', ...
- matlabroot,'/toolbox/compiler:', ...
- matlabroot,'/toolbox/control/control:', ...
- matlabroot,'/toolbox/control/ctrlguis:', ...
- matlabroot,'/toolbox/control/ctrlobsolete:', ...
- matlabroot,'/toolbox/control/ctrlutil:', ...
- matlabroot,'/toolbox/control/ctrldemos:', ...
- matlabroot,'/toolbox/shared/slcontrollib:', ...
- matlabroot,'/toolbox/curvefit/curvefit:', ...
- matlabroot,'/toolbox/curvefit/cftoolgui:', ...
- matlabroot,'/toolbox/curvefit/sftoolgui:', ...
- matlabroot,'/toolbox/shared/optimlib:', ...
- matlabroot,'/toolbox/database/database:', ...
- matlabroot,'/toolbox/database/dbdemos:', ...
- matlabroot,'/toolbox/database/vqb:', ...
- matlabroot,'/toolbox/datafeed/datafeed:', ...
- matlabroot,'/toolbox/datafeed/dfgui:', ...
- matlabroot,'/toolbox/distcomp:', ...
- matlabroot,'/toolbox/distcomp/distcomp:', ...
- matlabroot,'/toolbox/distcomp/user:', ...
- matlabroot,'/toolbox/distcomp/mpi:', ...
- matlabroot,'/toolbox/distcomp/pctdemos:', ...
- matlabroot,'/toolbox/distcomp/parallel:', ...
- matlabroot,'/toolbox/distcomp/parallel/datafun:', ...
- matlabroot,'/toolbox/distcomp/parallel/datatypes:', ...
- matlabroot,'/toolbox/distcomp/parallel/elfun:', ...
- matlabroot,'/toolbox/distcomp/parallel/elmat:', ...
- matlabroot,'/toolbox/distcomp/parallel/lapack:', ...
- matlabroot,'/toolbox/distcomp/parallel/matfun:', ...
- matlabroot,'/toolbox/distcomp/parallel/ops:', ...
- matlabroot,'/toolbox/distcomp/parallel/sparfun:', ...
- matlabroot,'/toolbox/distcomp/parallel/specfun:', ...
- matlabroot,'/toolbox/distcomp/parallel/util:', ...
- matlabroot,'/toolbox/distcomp/lang:', ...
- matlabroot,'/toolbox/dspblks/dspblks:', ...
- matlabroot,'/toolbox/dspblks/dspmasks:', ...
- matlabroot,'/toolbox/dspblks/dspmex:', ...
- matlabroot,'/toolbox/dspblks/dspdemos:', ...
- matlabroot,'/toolbox/shared/filterdesignlib:', ...
- matlabroot,'/help/toolbox/dspblks/examples:', ...
- matlabroot,'/toolbox/econ/econ:', ...
- matlabroot,'/toolbox/econ/econdemos:', ...
- matlabroot,'/toolbox/eml/eml:', ...
- matlabroot,'/toolbox/emlcoder/emlcoder:', ...
- matlabroot,'/toolbox/emlcoder/emlcodermex:', ...
- matlabroot,'/toolbox/shared/simtargets:', ...
- matlabroot,'/toolbox/finance/finance:', ...
- matlabroot,'/toolbox/finance/calendar:', ...
- matlabroot,'/toolbox/finance/findemos:', ...
- matlabroot,'/toolbox/finance/finsupport:', ...
- matlabroot,'/toolbox/finance/ftseries:', ...
- matlabroot,'/toolbox/fixedpoint/fixedpoint:', ...
- matlabroot,'/toolbox/fixedpoint/fidemos:', ...
- matlabroot,'/toolbox/fixedpoint/fixedpointtool:', ...
- matlabroot,'/toolbox/gads:', ...
- matlabroot,'/toolbox/gads/gads:', ...
- matlabroot,'/toolbox/gads/gadsdemos:', ...
- matlabroot,'/toolbox/ident/ident:', ...
- matlabroot,'/toolbox/ident/nlident:', ...
- matlabroot,'/toolbox/ident/idobsolete:', ...
- matlabroot,'/toolbox/ident/idguis:', ...
- matlabroot,'/toolbox/ident/idutils:', ...
- matlabroot,'/toolbox/ident/iddemos:', ...
- matlabroot,'/toolbox/ident/iddemos/examples:', ...
- matlabroot,'/toolbox/ident/idhelp:', ...
- matlabroot,'/toolbox/images/colorspaces:', ...
- matlabroot,'/toolbox/images/images:', ...
- matlabroot,'/toolbox/images/imdemos:', ...
- matlabroot,'/toolbox/images/imuitools:', ...
- matlabroot,'/toolbox/images/iptformats:', ...
- matlabroot,'/toolbox/images/iptutils:', ...
- matlabroot,'/toolbox/shared/imageslib:', ...
- matlabroot,'/toolbox/shared/spcuilib:', ...
- matlabroot,'/toolbox/instrument/instrument:', ...
- matlabroot,'/toolbox/instrument/instrumentdemos:', ...
- matlabroot,'/toolbox/instrument/instrumentblks/instrumentblks:', ...
- matlabroot,'/toolbox/instrument/instrumentblks/instrumentmex:', ...
- matlabroot,'/toolbox/instrument/instrumentblks/instrumentmasks:', ...
- matlabroot,'/toolbox/shared/testmeaslib:', ...
- matlabroot,'/toolbox/slvnv/simcoverage:', ...
- matlabroot,'/toolbox/nnet:', ...
- matlabroot,'/toolbox/nnet/nncontrol:', ...
- matlabroot,'/toolbox/nnet/nndemos:', ...
- matlabroot,'/toolbox/nnet/nnet:', ...
- matlabroot,'/toolbox/nnet/nnet/nnanalyze:', ...
- matlabroot,'/toolbox/nnet/nnet/nncustom:', ...
- matlabroot,'/toolbox/nnet/nnet/nndistance:', ...
- matlabroot,'/toolbox/nnet/nnet/nnformat:', ...
- matlabroot,'/toolbox/nnet/nnet/nninit:', ...
- matlabroot,'/toolbox/nnet/nnet/nnlearn:', ...
- matlabroot,'/toolbox/nnet/nnet/nnnetinput:', ...
- matlabroot,'/toolbox/nnet/nnet/nnnetwork:', ...
- matlabroot,'/toolbox/nnet/nnet/nnperformance:', ...
- matlabroot,'/toolbox/nnet/nnet/nnplot:', ...
- matlabroot,'/toolbox/nnet/nnet/nnprocess:', ...
- matlabroot,'/toolbox/nnet/nnet/nnsearch:', ...
- matlabroot,'/toolbox/nnet/nnet/nntopology:', ...
- matlabroot,'/toolbox/nnet/nnet/nntrain:', ...
- matlabroot,'/toolbox/nnet/nnet/nntransfer:', ...
- matlabroot,'/toolbox/nnet/nnet/nnweight:', ...
- matlabroot,'/toolbox/nnet/nnguis:', ...
- matlabroot,'/toolbox/nnet/nnguis/nftool:', ...
- matlabroot,'/toolbox/nnet/nnguis/nntool:', ...
- matlabroot,'/toolbox/nnet/nnobsolete:', ...
- matlabroot,'/toolbox/nnet/nnresource:', ...
- matlabroot,'/toolbox/nnet/nnutils:', ...
- matlabroot,'/toolbox/nnet/nnguis/nnguiutils:', ...
- matlabroot,'/toolbox/nnet/nndemos/nndatasets:', ...
- matlabroot,'/toolbox/nnet/nnguis/nntraintool:', ...
- matlabroot,'/toolbox/optim/optim:', ...
- matlabroot,'/toolbox/optim/optimdemos:', ...
- matlabroot,'/toolbox/pde:', ...
- matlabroot,'/toolbox/physmod/data_manager/data_manager:', ...
- matlabroot,'/toolbox/physmod/equation_language/equation_language:', ...
- matlabroot,'/toolbox/physmod/foundation/foundation:', ...
- matlabroot,'/toolbox/physmod/mech/importer:', ...
- matlabroot,'/toolbox/physmod/mech/mech:', ...
- matlabroot,'/toolbox/physmod/mech/mechdemos:', ...
- matlabroot,'/toolbox/physmod/network_engine/ne_sli:', ...
- matlabroot,'/toolbox/physmod/network_engine/ne_support:', ...
- matlabroot,'/toolbox/physmod/network_engine/network_engine:', ...
- matlabroot,'/toolbox/physmod/pm_sli/pm_sli:', ...
- matlabroot,'/toolbox/physmod/pm_visimpl/pm_visimpl:', ...
- matlabroot,'/toolbox/physmod/pmir/pmir:', ...
- matlabroot,'/toolbox/physmod/simscape/simscape:', ...
- matlabroot,'/toolbox/physmod/simscape/simscapedemos:', ...
- matlabroot,'/toolbox/physmod/simscape_language/simscape_language:', ...
- matlabroot,'/toolbox/physmod/unit_manager/unit_manager:', ...
- matlabroot,'/toolbox/signal/signal:', ...
- matlabroot,'/toolbox/signal/sigtools:', ...
- matlabroot,'/toolbox/signal/sptoolgui:', ...
- matlabroot,'/toolbox/signal/sigdemos:', ...
- matlabroot,'/toolbox/shared/siglib:', ...
- matlabroot,'/toolbox/sl3d/sl3d:', ...
- matlabroot,'/toolbox/sl3d/sl3ddemos:', ...
- matlabroot,'/toolbox/shared/sldv:', ...
- matlabroot,'/toolbox/splines:', ...
- matlabroot,'/toolbox/stats:', ...
- matlabroot,'/toolbox/shared/statslib:', ...
- matlabroot,'/toolbox/symbolic:', ...
- matlabroot,'/toolbox/wavelet/wavelet:', ...
- matlabroot,'/toolbox/wavelet/wmultisig1d:', ...
- matlabroot,'/toolbox/wavelet/wavedemo:', ...
- matlabroot,'/toolbox/wavelet/compression:', ...
-%%% END ENTRIES %%%
- ...
-];
-
-p = [userpath,p];
diff --git a/MatlabFiles/startd.m b/MatlabFiles/startd.m
index a741fef8f74533d6ecd29f5f10c32b245f823a31..71dfe3b09d4c5c25cd2da194a194aa0073343d6d 100755
--- a/MatlabFiles/startd.m
+++ b/MatlabFiles/startd.m
@@ -4,4 +4,4 @@ function startd(sd)
% use this function. E.g.
% sd=cd %to set sd to the current directory
% startd(sd)
-save '/Users/tzha/ZhaData/TZCcode/cstz/startdir0' sd
+save '/Users/tzha/ZhaData/Git/TZcode/MatlabFiles/startdir0' sd
diff --git a/MatlabFiles/startup.m b/MatlabFiles/startup.m
index a052b4ed4ea114ee9d2eeb53513f2e91f7c1e4e7..f7aeaa43f77fee77329443ac144eda0c8097238d 100755
--- a/MatlabFiles/startup.m
+++ b/MatlabFiles/startup.m
@@ -13,10 +13,10 @@ function startup()
%path(path,'d:\matlabr12\toolbox\cstz\lzpaper2')
%path(path,'d:\matlabr12\toolbox\cstz\rvarcode')
%path(path,'C:\Program Files\MATLAB\R2006b\work\cstz')
-path(path,'/Users/tzha/ZhaData/TZCcode/cstz')
-path(path,'/Users/tzha/ZhaData/TZCcode/cstz/MSV')
+path(path,'/Users/tzha/ZhaData/Git/TZcode/MatlabFiles')
+path(path,'/Users/tzha/ZhaData/Git/TZcode/MatlabFiles/MSV')
if exist('startdir0.mat')==2
- load /Users/tzha/ZhaData/TZCcode/cstz/startdir0
+ load /Users/tzha/ZhaData/Git/TZcode/MatlabFiles/startdir0
cd(sd)
end
format compact
diff --git a/tzDocumentation/Template_datainp_Markov.prn b/tzDocumentation/Template_datainp_Markov.prn
new file mode 100755
index 0000000000000000000000000000000000000000..9bc527183d009eab27e95f000b54b32c891d3a50
--- /dev/null
+++ b/tzDocumentation/Template_datainp_Markov.prn
@@ -0,0 +1,198 @@
+/******************************************************************************/
+/********************* Markov State Variable Information **********************/
+/******************************************************************************/
+
+//== Flat Independent Markov States and Simple Restrictions ==//
+
+
+//1st state variable: learning gain.
+//2nd state variable: average seigniorage.
+//3rd state variable: seigniorage shocks.
+//4th state variable: after-escape shocks.
+//+
+//Case: tv_varibles == (TV_GAIN | TV_AD | TV_DSHOCKS) && syntypes == S_D_SHOCKS. See datainp_setup.prn.
+//== Number Independent State Variables ==//
+2
+
+//=====================================================//
+//== state_variable[i] (1 <= i <= n_state_variables) ==//
+//=====================================================//
+//== Number of states for state_variable[1] ==//
+2
+
+
+//----------------- Prior for the transition matrix Q_k -----------------
+//== Each column contains the parameters for a Dirichlet prior on the corresponding
+//== column of the transition matrix. Each element must be positive. For each column,
+//== the relative size of the prior elements determine the relative size of the elements
+//== of the transition matrix and overall larger sizes implies a tighter prior.
+//If it is an identity matrix, we have no prior (i.e., a flat prior in this case).
+//Examples:
+//2.000000 1.000000
+//1.000000 2.000000
+//2.000000 1.000000 1.000000
+//1.000000 2.000000 1.000000
+//1.000000 1.000000 2.000000
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+1.000000 1.000000
+1.000000 1.000000
+
+
+//----------------- The semi-free parameter vector free -----------------
+//The vector, free, indictates the number of semi-free parameters in each column of Q_k. By semi-free, we
+// meen that the sum of each column in Q_k is 1.0 so that the number of true free parameters is one less.
+//== Free Dirichet dimensions for state_variable[1] ==//
+2 2
+
+
+//----------------- The restriction matrix R_j -----------------
+//== The jth restriction matrix is n_states x free[j]. Each row of the restriction
+//== matrix has exactly one non-zero entry and the sum of each column must be one.
+//Example 1:
+//1 0 0
+//0 1 0
+//0 0 1
+//Example 2 (3 states but 2 semi-free parameters for column j of Q_1):
+//0 0
+//1 0
+//0 1
+//which gives the free-parameter vector [0 q_2 q_3] where q_2 + q_3 = 1.0.
+//== Column restrictions for state_variable[1] ==//
+1 0
+0 1
+
+1 0
+0 1
+
+
+//----------------- Allows for lagged values of the state variable to be encoded -----------------
+//== Number of lags encoded for state_variable[1] ==//
+1
+
+
+//== Number of states for state_variable[2] ==//
+3
+
+//== Each column contains the parameters for a Dirichlet prior on the corresponding
+//== column of the transition matrix. Each element must be positive. For each column,
+//== the relative size of the prior elements determine the relative size of the elements
+//== of the transition matrix and overall larger sizes implies a tighter prior.
+//== Transition matrix prior for state_variable[2]. (n_states x n_states) ==//
+ 1.0 1.0 1.0
+ 1.0 1.0 1.0
+ 1.0 1.0 1.0
+
+//== Free Dirichet dimensions for state_variable[2] ==//
+2 2 2
+
+//== The jth restriction matrix is n_states x free[j]. Each row of the restriction
+//== matrix has exactly one non-zero entry and the sum of each column must be one.
+//== Column restrictions for state_variable[2] ==//
+1 0
+0 0
+0 1
+
+0 0
+1 0
+0 1
+
+0 0.5
+0 0.5
+1 0
+
+
+//----------------- Allows for lagged values of the state variable to be encoded -----------------
+//== Number of lags encoded for state_variable[1] ==//
+0
+
+
+//----------------- User-defined intial values of the transition matrix Q_k, mainly for debugging purposes -----------------
+//Non-zero integer: use the values under Initial: Transition matrix; 0: use Waggoner's default value which is the prior mean.
+//== indxInitializeTransitionMatrix ==//
+1
+
+
+// 0.5 0.5
+// 0.5 0.5
+//Call ReadTransitionMatrices(f,(char*)NULL,"Initial: ",model) in the main function.
+//If indxInitializeTransitionMatrix=0 or does not exist, the following values are not used and
+// Waggoner's function initializes the value from the prior mean.
+//== Initial: Transition matrix[1] ==//
+ 9.7331908462961059e-01 1.5790335972902768e-02
+ 2.6680915370389413e-02 9.8420966402709720e-01
+
+
+
+//--- Initial guess
+ 9.7331908462961059e-01 1.5790335972902768e-02
+ 2.6680915370389413e-02 9.8420966402709720e-01
+//--- Converged results
+ 9.7885175490925935e-001 1.6213439424811054e-002
+ 2.1148245090740647e-002 9.8378656057518898e-001
+
+
+
+
+//== Initial: Transition matrix[2] ==//
+ 3.8872218833300276e-01 0.0 3.185699602083008e-002
+ 0.0 9.8480151496509816e-01 3.185699602083008e-002
+ 6.112778116669972e-001 1.519848503490184e-002 9.3628600795833983e-01
+
+
+//--- Initial guess
+ 3.8872218833300276e-01 0.0 3.185699602083008e-002
+ 0.0 9.8480151496509816e-01 3.185699602083008e-002
+ 6.112778116669972e-001 1.519848503490184e-002 9.3628600795833983e-01
+//--- Converged results
+ 4.3952235591344235e-001 0.0000000000000000e+000 3.6997582827338338e-002
+ 0.0000000000000000e+000 9.7128777689293189e-001 3.6997582827338338e-002
+ 5.6047764408655765e-001 2.8712223107068113e-002 9.2600483434532332e-001
+
+
+
+
+
+
+
+
+
+
+
+
+//********************************** Notes ******************************************//
+//----------------- Not working any more but used to have the user-defined intial values of Q_k -----------------
+//== Initial transition matrix for state_variable[12]. (n_states x n_states) ==//
+0.99 0.01
+0.01 0.99
+
+The following no longer used in the new switch.c:
+line 564 switch.c.
+// Initial transition matrix[1] //
+0.8 0.1 0.1
+0.1 0.8 0.1
+0.1 0.1 0.8
+
+
+For the restrictions of the absorbing type:
+X X 0
+X X 0
+X X 1
+
+we have:
+
+// Free parameters in state_variable[1] //
+3 3 1
+
+// Column restrictions for state_variable[1] //
+1 0 0
+0 1 0
+0 0 1
+
+1 0 0
+0 1 0
+0 0 1
+
+0
+0
+1
+//--------------------------------------------------------------------------//
diff --git a/tzDocumentation/readme_use_dwVARestimate.prn b/tzDocumentation/readme_use_dwVARestimate.prn
new file mode 100755
index 0000000000000000000000000000000000000000..8dec4271077fcc09828f036a94a8082b2ad1fd4c
--- /dev/null
+++ b/tzDocumentation/readme_use_dwVARestimate.prn
@@ -0,0 +1,166 @@
+Corresponding to the SWZ article.
+----------------------------------
+Zeta in DW's output is Xi^2 in SWZ's JE article. The prior on Zeta is gamma(a=1,b=1). <<>> Check with Dan to be sure (17 Jan 2010 with Kirstin)
+ Sigma(k) = inv( A0(K) Zeta(k) A0(k)' )
+
+
+
+In the init_*.dat, the following is always true (hard-coded when using create_init_file.c):
+//== Specification (0=default 1=Sims-Zha 2=Random Walk) ==//
+1
+
+But if you want to change this, you must go into this file to do it manually.
+
+
+
+3by2 (e.g., 3 states for variances and 2 states for coefficients)
+
+sv 1 sv 2
+3 states 2 states
+A1-A3 B1-B2
+
+Block A1
+B1 and B2
+
+Block A2
+B1 and B2
+
+Block A3
+B1 and B2
+==>
+Overall Sates: Individual Sates
+1. A1 B1
+2. A1 B2
+3. A2 B1
+4. A2 B2
+5. A3 B1
+6. A3 B2
+
+
+WriteTransitionMatrices(f_out,(char*)NULL,header,model);
+Write_VAR_Parameters(f_out,(char*)NULL,header,model);
+
+
+create_init_file <
+TZMatlab file> <DWSpecification file> file_tag name.
+Example
+exectuable datainp2_a_case2p2_sz5v_lp5011_d8507.prn specification2p2_3by2.dat test
+All these do is to read the SZ prior and the data and the starting point for the estimation.
+
+In Specification*.dat:
+//== Controlling states variables for coefficients ==//
+
+//== Controlling states variables for variance ==//
+
+
+In init_tag.dat, we have
+
+Mean of the prior.
+//== Initial: Transition matrix[1] ==//
+
+//== Initial: Transition matrix[2] ==//
+
+
+
+
+If -ft is chosen, -fs, -fr, and -rh will be all ignored.
+For -fr option, supply a different file with the data from //== Initial: Transition matrix[1] ==// on.
+ Perhaps change to //== Restart: ?????==//
+-rh "Restart: " in the command line.
+ or if I'm lazy, I would do the following instead:
+-rh "Initial: " in the command line.
+
+Another example:
+
+executable -ft tag -MLE -ce 1.03-6 -ii 1.5
+ //-ii can be a double.
+
+
+
+
+
+
+Input Command line:
+
+Group 1 (must specify most inputs).
+ Attempt to set up model from command line. Command line options are the following
+
+ -di <directory>
+ If this argument exists, then all input files are in specified directory.
+
+ -ft <filename tag>
+ If this argument exists, then the following is attempted:
+
+ specification file name: est_final_<tag>.dat
+ init/restart file name: est_final_<tag>.dat with header="Posterior mode: "
+
+ specification file name: init_<tag>.dat
+ init/restart file name: est_intermediate_<tag>.dat with header="Iteration %d: "
+
+ (not yet implemented)
+ specification file name: init_<tag>.dat
+ init/restart file name: est_csminwel_<tag>.dat
+
+ specification file name: init_<tag>.dat
+ init/restart file name: init_<tag>.dat with header="Initial: "
+
+ Failure to load both the specification and restart/init files causes the routine to exit.
+
+ -fs <filename>
+ If this argument exists, then the specification file name is <filename>. The argument -ft
+ takes precedence over -fs.
+
+ -fr <filename>
+ If this argument exists, then the init/restart file name is <filename>. Must be used in
+ conjunction with the argument -fs. The default value is the filename associated with the
+ argument -fs.
+
+ -rh <header>
+ If this argument exists, then the header for the init/restart file is <header>. Must be
+ used in conjuction with the arguments -fr or -fs. The default value is "".
+
+ If no command line options are given, then attemps to use a default input file
+ with the name "default.ini". Returns one valid pointer to a TStateModel upon
+ success and null upon failure.
+
+
+Group 2 (either output or controlling execution of csminwel) (has a lot of default values).
+ Attempt to set up model from command line. Command line options are the following
+
+ -do <directory>
+ If this argument exists, then all output files are put in the specified directory.
+
+ -fo <filename tag>
+ If this argument exists, then the output files are
+
+ est_csminwel_<tag>.dat
+ est_intermediate_<tag>.dat
+ est_final_<tag>.dat
+
+ The default value is the filename tag associated with the argument -ft if it exists. Otherwise
+ it is "default".
+
+ -MLE
+ Find the maximum likelihood estimate
+
+ -PM (default)
+ Find the posterior mode
+
+ -cb <floating point number> (default = 1.0e-3)
+ Beginning csminwel exit criterion
+
+ -ce <floating point number> (default = 1.03-6)
+ Ending csminwel exit criterion
+
+ -ci <floating point number> (default = 0.1)
+ csminwel exit criterion increment multiplier
+
+ -ib <integer> (default = 50)
+ Beginning csminwel maximum iteration value
+
+ -ii <floating point number> (default = 2)
+ csminwel maximum interation increment multiplier
+
+ If no command line options are given, then attemps to use a default input file
+ with the name "default.ini". Returns one valid pointer to a TStateModel upon
+ success and null upon failure.
diff --git a/tzDocumentation/readme_use_dwVARmhm.prn b/tzDocumentation/readme_use_dwVARmhm.prn
new file mode 100755
index 0000000000000000000000000000000000000000..2132a3bd5b15d9464a7fe881afedf4afcd4f0c4e
--- /dev/null
+++ b/tzDocumentation/readme_use_dwVARmhm.prn
@@ -0,0 +1,49 @@
+Example:
+
+swzmnhstage1 -fi MHM_input.dat -ft sz5v_2vby2m
+
+Command line:
+
+ Attempt to set up model from command line. Command line options are the
+ following
+
+ -di <directory>
+ If this argument exists, then all input files are in specified directory.
+
+ -do <directory>
+ If this argument exists, then all output files are in specified directory.
+
+ -ft <filename tag>
+ If this argument exists, then the following is attempted:
+
+ 1) specification file name: mhm_final_<tag>.dat
+ mhm arguments file name: mhm_final_<tag>.dat
+
+ 2) specification file name: mhm_intermediate_<tag>.dat
+ mhm arguments file name: mhm_intermediate_<tag>.dat
+
+ 3) specification file name: est_final_<tag>.dat
+ mhm arguments file name: -fi <filename>
+
+ -fi <filename>
+ If this argument exists, then additional mhm arguments are read from the
+ input file with the given filename.
+
+ -fs <filename>
+ If this argument exists, then the specification file name is <filename>.
+ The argument -ft takes precedence over -fs.
+
+ -fp <filename>
+ If this argument exists, then the posterior is read from <filename>. Must
+ be used in conjunction with the argument -fs. The default value is the
+ filename associated with the argument -fs.
+
+ -ph <header>
+ If this argument exists, then the header for the posterior file is
+ <header>. Must be used in conjuction with the arguments -fp or -fs. The
+ default value is "Posterior mode: ".
+
+ If no command line options are given, then attemps to use a default input file
+ with the name "default.ini". Returns one valid pointer to a TStateModel upon
+ success and null upon failure.
+
diff --git a/tzDocumentation/readme_use_dwswitch.prn b/tzDocumentation/readme_use_dwswitch.prn
new file mode 100755
index 0000000000000000000000000000000000000000..71c5d7e976bec2a9adba6ccb2151bd1aa122fd30
--- /dev/null
+++ b/tzDocumentation/readme_use_dwswitch.prn
@@ -0,0 +1,792 @@
+ ********* Contents **********
+ ********** Guide for using Waggoner switch.c and switch_opt.c. **********
+/* Basic steps
+/* How to use DW regime contral table for DSGE and any regime-switching models
+/* Some basic DW's functions in his switch.h, such as likelihood, posterior, probabilities of regimes.
+/* Calling some switching variables from DW code
+ (1) Regime swaps.
+ (2) Grand, base, lagged regimes.
+ (3) Getting P(s_t = i | Y_{t-1}, theta) or P(s_t = i | Y_{t}, theta) using DW's ElementV() where
+ t is base-1 (1 to T) while i is base-0 (regimes 0 to h-1).
+ (4) Getting base regimes.
+ (5) Getting number of lags for a particular regime variable.
+ (6) Getting base transition probabilities.
+/* Notes on block-wise optimization in switch_opt.c
+/* How are overall states (regimes) ordered
+/* Dirichlet prior for a column of the transition probability matrix
+ in DW's initial Markov-switching data file used by the SWZ estimation procedure
+/* How to specify time variations for coefficients and variance sizes for the VAR model
+/* How to convert between model free parameters and model specific parameters (no transition matrix Q is involved)
+/* How to specify restrictions on each transition matrix Q_k
+/* How to obtain starting point of the transition matrix Q_k
+/* How to reset the probability of the initial state s_0
+/* How to set the transition matrix for the 3rd state that is redundant to the 2nd state
+/* How to call base regimes
+/* How to write out transition matrices
+/*
+/* DSGE output files such as simulations and impulse responses
+ *****************************
+
+
+//==================================
+// Basic steps
+//==================================
+(-1) In extern_dw_switch.c, toggel between the old DW switch program
+ #include "switch.c"
+ #include "switchio.c"
+ #include "switch_opt.c"
+ and the new DW switch program
+ #include "dsge_switch.c"
+ #include "dsge_switchio.c"
+ #include "dsge_switch_opt.c"
+ #include "Dirichlet.c"
+
+(0) For the new DW switch program, make sure that in tzmatlab.h, use
+ #define NEWVERSIONofDW_SWITCH
+
+(1) In matrix.h, make sure the following line is activated and the other related lines (e.g., STRUCTURED_COLUMN_MAJOR) are disactivated:
+ #define TZ_COLUMN_MAJOR
+
+ In general toggle between #define STRUCTURED_COLUMN_MAJOR and #define TZ_COLUMN_MAJOR;
+
+(2) In matrix.h, make sure the following line is activated and the line #include "tz2dw.h" is disactivated:
+ #include "tzmatlab.h"
+
+ In general toggle between #include "tzmatlab.h" and #include "tz2dw.h"
+
+(3) In bmatrix.c, always comment out the following for Windows (Linux seems working fine; if not, we can always comment it out).
+
+ /********************
+ #include "blas_lapack.h"
+ #define USE_BLAS_LAPACK
+ /********************/
+
+ When this is commented out, Dan's code will work slower because it does not use any Blas and Lapack calls,
+ but Tao's code will be fine because he is using MKL.
+
+(4) To use the old code, uncomment the function in switch.c (at the end of this file):
+ TParameters* CreateParameters()
+
+(5) Make sure to uncomment out the function
+ int main(void)
+ in switchio.c.
+
+
+//==================================
+// How to use DW regime contral table for DSGE and any regime-switching models
+//==================================
+There three blocks of parameters: constant, coefficient TV, and volatility TV.
+It is better that each block is NOT empty. If for some reasons one of the blocks is empty, saying,
+ the coeffivient TV block (Time varying coefficient parameters) is set to 0, then
+ use the following convention:
+ DWRegimeControlVariables_ps->cum_total_coef in the DW code will be NULL;
+ DWRegimeControlVariables_ps->indx_st_coef in the DW code will be NULL;
+ In this case, do not call dw_DimA(n_st_coef) because the program will crash and give an error.
+ To make it automatic, we can do this: if it is NULL, do not call dw_DimA(n_st_coef) or
+ create an macro (use define) to globally return (define) dw_DimA(n_st_coef) to be 0.
+
+
+Illustration:
+
+Parameters not time varying
+------
+X
+X
+X
+
+Coefficient parameters time varying
+------
+1st coefficient (2 regimes)
+X <-- cum_total_coef[0] (=3)
+X
+
+2nd coefficient (3 regimes)
+X <-- cum_total_coef[1] (=5)
+X
+X <-- cum_total_coef[1] + (n_st_coef[1]-1) (= 7)
+
+Variance parameters time varying
+------
+1st variance parameter (2 regimes)
+X <-- cum_total_var[0] (=8)
+X
+
+2st variance parameter (1 regime)
+X <-- cum_total_var[1] (=10)
+
+3nd variance parameter (3 regimes)
+X
+X
+X
+
+Notes: (a) dw_DimA(n_st_coef) = 2 (coef parameters), dw_DimA(n_st_var) = 3 (variance parameters);
+ (b) n_st_coef[0] = 2 (regimes), n_st_coef[1] = 3 (regimes);
+ n_st_var[0] = 2 (regimes), n_st_var[1] = 1 (regime), n_st_var[2] = 3 (regimes);
+
+
+typedef struct TSDWRegimeControlVariables_tag
+{
+ //--- Memory of all the following arrays will be allocated by DW and destroyed by the DW function dw_FreeArray(),
+ //--- partly because DW allocated array with some information stored in the -2, -1 elements, so
+ //--- the array does not begin with base 0. Thus, tzDestroy(n_coef) or free(n_coef)will NOT work.
+ //--- Thus, it's essential to destroy them using dw_FreeArray().
+
+ int n_constpars; // number of constant (not time varying) free parameters.
+
+ int *n_st_coef; //n_st_coef[i-1] is the number of regimes associated with the i_th time-varying coefficient parameter.
+ int *cum_total_coef; //cum_total_coef[i-1] is the offset of the i_th time-varying coefficient parameter
+ // in the whole vector of free parameters.
+ int **indx_st_coef; //indx_st_coef[st][i-1] is the regime associated with the i_th time-varying
+ // variance parameter with the grand regime st.
+
+ int *n_st_var; //n_st_var[i-1] is the number of regimes associated with the i_th time-varying variance parameter.
+ int *cum_total_var; //cum_total_var[i-1] is the offset of the i_th time-varying variance parameter
+ // in the whole vector of free parameters.
+ int **indx_st_var; //indx_var[st][i-1] is the regime associated with the i_th time-varying
+ // variance parameter with the grand regime st.
+ //Use 1: g_sigma_w_dv->v[indx_st_var[st][2]] where 2 means that g_sigma_w_dv is the 3rd
+ // time-varying variance parameter to be considered.
+ //Use 2: xphi_dv->v[cum_total_var[2]+indx_st_var[st][2]] where 2 means the 3rd time-varying
+ // variance parameters; this value is the same as g_sigma_w_dv->v[idx_var[st][2]].
+} TSDWRegimeControlVariables;
+
+
+
+//==================================
+// Some basic DW's functions in his switch.h, such as likelihood, posterior, probabilities of regimes.
+//==================================
+// ln(P(y[t] | Y[t-1], Z[t], s[t], theta, q))
+#define LogConditionalLikelihood(s,t,model) ((model)->routines->pLogConditionalLikelihood(s,t,model))
+
+// ln(P(y[t] | Y[t-1], Z[t], theta, q))
+PRECISION LogConditionalLikelihood_StatesIntegratedOut(int t, TStateModel *model);
+
+// E[y[t] | Y[t-1], Z[t], s[t], theta, q]
+#define ExpectationSingleStep(y,s,t,model) ((model)->routines->pExpectationSingleStep(y,s,t,model))
+
+// E[y[t] | Y[t-1], Z[t], theta, q]
+TVector ExpectationSingleStep_StatesIntegratedOut(TVector y, int t, TStateModel *model);
+
+// ln(P(q))
+PRECISION LogPrior_q(TStateModel *model);
+
+// ln(P(theta))
+#define LogPrior_theta(model) ((model)->routines->pLogPrior(model))
+
+// ln(P(theta, q))
+#define LogPrior(model) (LogPrior_theta(model) + LogPrior_q(model))
+
+// ln(P(S[T] | theta, q))
+PRECISION LogConditionalPrior_S(TStateModel *model);
+
+// ln(P(Y[T] | Z[T], S[T], theta, q)) = Assumes that the data is for fobs <= t <= nobs
+PRECISION LogLikelihood(TStateModel *model);
+
+// ln(P(Y[T] | Z[T], S[T], theta, Q) * P(S[T] | theta, q) * P(theta, q))
+#define LogPosterior(model) (LogLikelihood(model) + LogConditionalPrior_S(model) + LogPrior(model))
+
+// ln(P(Y[T] | Z[T], theta, Q)) - Assumes that the data is for fobs <= t <= nobs
+PRECISION LogLikelihood_StatesIntegratedOut(TStateModel *model);
+
+// ln(P(Y[T] | Z[T], theta, q) * P(Theta, q))
+#define LogPosterior_StatesIntegratedOut(model) (LogLikelihood_StatesIntegratedOut(model) + LogPrior(model))
+
+// ln(P(S[T] = S | Y[T], Z[T], theta, q))
+PRECISION LogConditionalProbabilityStates(int *S, TStateModel *model);
+
+// P(s[t] = s | Y[t], Z[t], theta, q)
+PRECISION ProbabilityStateConditionalCurrent(int s, int t, TStateModel *model);
+
+// P(s[t] = s | Y[t-1], Z[t-1], theta, q)
+PRECISION ProbabilityStateConditionalPrevious(int s, int t, TStateModel *model);
+
+// p[t] = P(s[t] = s | Y[T], Z[T], theta, q)
+TVector ProbabilitiesState(TVector p, int s, TStateModel *model);
+
+
+
+//==================================
+// Calling some switching variables from DW code
+// (1) Regime swaps.
+// (2) Grand, base, lagged regimes.
+// (3) Getting P(s_t = i | Y_{t-1}, theta) or P(s_t = i | Y_{t}, theta) using DW's ElementV() where
+// t is base-1 (1 to T) while i is base-0 (regimes 0 to h-1).
+// (4) Getting base regimes.
+// (5) Getting number of lags for a particular regime variable.
+// (6) Getting base transition probabilities.
+//==================================
+(1) Regime swaps.
+ Swap_SV(smodel_ps->sv->state_variable[0], 0, si);
+
+
+(2) Grand, base, lagged regimes.
+ //------------ Index individual states from the grand state. -------------------//
+ Define the following:
+ struct TStateModel_tag *smodel
+ struct TMarkovStateVariable_tag *sv
+
+ smodel->sv->nstates (***grand*** number of states, seldom used by the user)
+ smodel->sv->n_state_variables (number of state variables)
+ smodel->sv->state_variable[0]->nbasestates (number of states for the 1st state variable)
+ smodel->sv->state_variable[0]->nstates (number of ***sub-grand** states for the 1st state variable,
+ which equals nbasestates^nlags), where nlags is specified in the Markov input data file (see,
+ for example, template_datainp_Markov.prn) as in the following 1-lag case for the 1st state variable:
+
+ //-----------------------------------------------------------------------------//
+ //-- Allows for lagged values of the state variable to be encoded --//
+ //-----------------------------------------------------------------------------//
+ //== Number of lags encoded for state_variable[1] ==//
+ 1
+
+ smodel->sv->state_variable[0]->lag_index (lagged states for the 1st state variable)
+ smodel->sv->state_variable[0]->nlags_encoded (number of lags of the states for the 1st state variable)
+ smodel->sv->state_variable[0]->Q (transition matrix for 1st state variable)
+ Equivalently, smodel->sv->QA[0];
+
+
+ In the file switch.h, there is a structure called TMarkovStateVariable.
+ This sturcture contains a field called lag_index which is an array of arrays of integers.
+ This field provides for decoding the state indexes. For instance, suppose there are 2 lags encoded,
+ so the grand state k would consist of three states (i_0,i_1,i_2) which would be the current state,
+ the first lagged state, and the second lagged state. In this case we have that
+
+ lag_index[k][0] = i_0
+ lag_index[k][1] = i_1
+ lag_index[k][2] = i_2
+
+ This allows you to easily decode the grand state k. Currently, Dan does not have a procedure for
+ going in the other direction to encode the grand state from the current and lagged states.
+
+
+(3) Getting P(s_t = i | Y_{t-1}, theta) or P(s_t = i | Y_{t}, theta) using DW's ElementV() where
+ t is base-1 (1 to T) while i is base-0 (regimes 0 to h-1).
+ int _t; //base-1 where my own C data is base-0, thus we need to use _t-1.
+ int _i; //base-0 regimes at time t.
+ P(s_t = i | Y_{t-1}, theta) = ElementV(smodel_ps->Z[_t],_i);
+ P(s_t = i | Y_{t}, theta) = ElementV(smodel_ps->V[_t],_i);
+ //---------------- Real example (OLD Code) -----------------//
+ prob_gainstate_dv->v[si] = 0.0;
+ for (ki=smodel_ps->sv->nstates-1; ki>=0; ki--)
+ if (smodel_ps->sv->state_variable[0]->lag_index[Index[ki][0]][0]==si)
+ prob_gainstate_dv->v[si] += ElementV(smodel_ps->Z[inpt],ki);
+ //July/2007. prob_gainstate_dv->v[si] += ProbabilityStateGivenPreviousData(ki,inpt,smodel_ps);
+ //Cannot call the above function because it calls the likelihood function which creates an infinite loop.
+ gBeta_dm->M[mos(inpt,si,gBeta_dm->nrows)] = gBeta_dm->M[mloc] + prob_gainstate_dv->v[si]*exp(tvp7d_ps->loggain_dv->v[si]) * (ginf_dv->v[(nlags_max-1)+inpt0] - gBeta_dm->M[mloc]);
+
+
+(4) Getting base regimes.
+ Consider the following example
+ s_t = (m*_t, v_t) where m*_t = (m_t, m_{t-1}).
+ for s_t = k, we have
+ mstar = smodel_ps->sv->index[k][0]; //Old Code: smodel_ps->sv->Index[k][0];
+ v = smodel_ps->sv->index[k][1]; //Old Code: smodel_ps->sv->Index[k][1];
+ m_0 = smodel_ps->sv->state_variable[0]->lag_index[mstar][0];
+ m_1 = smodel_ps->sv->state_variable[0]->lag_index[mstar][1];
+ such that m_t = m_0, m_{t-1} = m_1, v_t = v.
+
+
+(5) Getting number of lags for a particular regime variable.
+ int nlags_statevariable0 = smodel_ps->sv->state_variable[0]->nlags_encoded;
+
+
+(6) Getting base transition probabilities.
+ New Code
+ -------------
+ sv_ComputeTransitionMatrix(0, smodel_ps->sv, smodel_ps); //0 mean time 0 (in the time-varying Q, t means time t)
+ //This line must be used before copying it to any Q in case theta or q or anything has changed.
+ CopyMatrix0(Qc_dm, smodel_ps->sv->state_variable[0]->baseQ); //$$##
+ Qc_dm->flag = M_GE;
+
+
+ OLD Code
+ -------------
+ Example: getting Qm for m_t.
+
+ Allocate the memory for MyMatrixQm and call
+
+ GetBaseTransitionMatrix_SV(MyMatrixQm, smodel->sv->state_variable[0]);
+
+ or call the following function, which creates MyMatrixQm as well
+
+ MyMatrixQm = GetBaseTransitionMatrix_SV(NULL, smodel->sv->state_variable[0]);
+
+ //--- Getting Qv for v_t ---
+ Simply call
+
+ smodel->sv->QA[1];
+
+ or
+
+ CopyMatrix0(MyMatrixQv, smodel->sv->QA[1]);
+
+ or
+
+ GetBaseTransitionMatrix_SV(MyMatrixQv, smodel->sv->state_variable[1]);
+
+
+
+//==================================
+// Notes on block-wise optimization in switch_opt.c
+//==================================
+In switch_opt.c, we have
+
+void SetupObjectiveFunction(TStateModel *model, PRECISION *Modified, PRECISION *FreeQ, PRECISION *FreeTheta)
+
+where FreeTheta points to free model parameters only (no free parameters in transition matricies);
+ FreeQ points to free q's in the transtiona matrices only;
+ Modified ponts to the location to be modified.
+*** Note that FreeQ and FreeTheta need not be adjacent. ***
+
+
+
+//==================================
+// How are overall states (regimes) ordered
+//==================================
+sv 1 sv 2
+3 states 2 states
+A1-A3 B1-B2
+
+Block A1
+B1 and B2
+
+Block A2
+B1 and B2
+
+Block A3
+B1 and B2
+==>
+Overall Sates: Individual Sates
+1. A1 B1
+2. A1 B2
+3. A2 B1
+4. A2 B2
+5. A3 B1
+6. A3 B2
+
+
+
+//==================================
+// Dirichlet prior for a column of the transition probability matrix
+// in DW's initial Markov-switching data file used by the SWZ estimation procedure
+//==================================
+Let x_h be a Dirichlet of form of Gelman (2004) where h be the number of states,
+ a_h be the diagonal parameter, and a_1, ..., a_{h-1}=1 to allow for an absorbing state.
+Let p be the probability of staying in that regime (for 20 months -- 1.7 years, p = 0.95;
+ for 6.7 quarters -- 1.67 years, p = 0.85)
+For E(x_h) = p, we have a_h/(h-1+a_h) = p, ==>
+ a_h = ( p/(1-p) )*(h-1)
+
+For monthly data, h=2, and p=0.95, a_h = 19.0e+000, ==>
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+ 19.0e+000 1.0000000000000000e+000
+ 1.0000000000000000e+000 19.0e+000
+
+For quarterly data, h=2, and p=0.85, a_h = 5.6666666666666661e+000, ==>
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+ 5.6666666666666661e+000 1.0000000000000000e+000
+ 1.0000000000000000e+000 5.6666666666666661e+000
+
+For quarterly data and restricted Q:
+ a (1-b)/2 0
+ 1-a b 1-c
+ 0 (1-b)/2 c
+h=3, and p=0.85, a_h = 5.67, ==>
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+ 5.67 1.0 1.0
+ 1.0 5.67 1.0
+ 1.0 1.0 5.67
+
+For monthly data and restricted Q:
+ a (1-b)/2 0
+ 1-a b 1-c
+ 0 (1-b)/2 c
+h=3, and p=0.95, a_h = 19.0, ==>
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+ 19.000 1.0 1.0
+ 1.0 19.00 1.0
+ 1.0 1.0 19.00
+
+
+For quarterly data and unrestricted Q, h=3, and p=0.85, a_h = 11.333333, ==>
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+ 11.33 1.0 1.0
+ 1.0 11.33 1.0
+ 1.0 1.0 11.33
+
+For quarterly data and unrestricted Q, h=4, and p=0.85, a_h = 17.0, ==>
+//== Transition matrix prior for state_variable[1]. (n_states x n_states) ==//
+ 17.0 1.0 1.0 1.0
+ 1.0 17.0 1.0 1.0
+ 1.0 1.0 17.0 1.0
+ 1.0 1.0 1.0 17.0
+
+
+
+//==================================
+// How to specify hard restrictions on each transitiona matrix Q_k
+// in DW's initial Markov-switching data file used by the SWZ estimation procedure
+//==================================
+In specification*.prn or datainp_markovswitching*.prn or any file that contains Markov-switching information,
+ follow the following template:
+
+Suppose Q = [
+ p1 (1-p2)/2 0
+ 1-p1 p2 1-p2
+ 0 (1-p2)/2 p2
+ ];
+We have
+
+//== Free Dirichet dimensions for state_variable[1] ==//
+2 2 2
+
+//== The jth restriction matrix is n_states-by-free[j]. Each row of the restriction
+//== matrix has exactly one non-zero entry and the sum of each column must be one.
+//== Column restrictions for state_variable[1] ==//
+1 0
+0 1
+0 0
+
+0.5 0
+0 1
+0.5 0
+
+0 0
+1 0
+0 1
+
+For the restrictions of the absorbing type:
+ X X 0
+ X X 0
+ X X 1
+we have:
+
+// Free parameters in state_variable[1] //
+3 3 1
+
+// Column restrictions for state_variable[1] //
+1 0 0
+0 1 0
+0 0 1
+
+1 0 0
+0 1 0
+0 0 1
+
+0
+0
+1
+
+Other examples:
+
+Example 1:
+1 0 0
+0 1 0
+0 0 1
+Example 2 (3 states but 2 semi-free parameters for column j of Q_1):
+0 0
+1 0
+0 1
+which gives the free-parameter vector [0 q_2 q_3] where q_2 + q_3 = 1.0.
+
+
+
+//==================================
+// How to specify time variations for coefficients and variance sizes for the VAR model
+//==================================
+Note that the first column corresponds to the first state variable, and so forth.
+
+For example:
+
+//== nvar x n_state_variables matrix. In the jth row, a non-zero value implies that
+//== this state variable controls the jth column of A0 and Aplus
+//== Controlling states variables for coefficients ==//
+0 0
+0 0
+0 0
+0 0
+0 1
+
+//== nvar x n_state_variables matrix. In the jth row, a non-zero value implies that
+//== this state variable controls the jth diagonal element of Xi
+//== Controlling states variables for variance ==//
+ 1 0
+ 1 0
+ 1 0
+ 1 0
+ 1 0
+
+Clearly, the 1st state variable corresponds to variance sizes and the 2nd state variable corresponds to coefficients.
+ for coefficient, only the fifth equation is time variation.
+
+
+
+
+//==================================
+// How to convert between model free parameters and model specific parameters (no transition matrix Q is involved)
+//==================================
+When calling Waggoner's function CreateParameters(), one needs supply the two functions to which Waggoner's
+ function names point. These two Waggoner's function names (of course, the functions are supplied by the
+ user himself) are:
+ ConvertFreeParametersToTheta(smodel_ps, xphi_sdv.v); //Waggoner's function.
+ //Converting the vectorized free model parameters and the model specific (and meaningful) parameters.
+ ConvertFreeParametersToQ(smodel_ps, xqs_sdv.v); //Waggoner's function.
+ //Converting the model specific (and meaningful) parameters to the vectorized free model parameters.
+
+
+
+
+//==================================
+// How to specify restrictions on each transitiona matrix Q_k
+//==================================
+//----------------- The semi-free parameter vector: free -----------------
+//The vector, free, indictates the number of semi-free parameters in each column of Q_k. By semi-free, we
+// meen that the sum of each column in Q_k is 1.0 so that the number of true free parameters is one less.
+//== Free Dirichet dimensions for state_variable[1] ==//
+2 2
+
+//----------------- The restriction matrix R_j -----------------
+//== The jth restriction matrix is n_states-by-free[j]. Each row of the restriction
+//== matrix has exactly one non-zero entry and the sum of each column must be one.
+
+
+
+
+
+
+//==================================
+// How to obtain starting point of the transition matrix Q_k
+//==================================
+Without supplying initial Q, DW's code will use the prior mean of Q as a starting point (default).
+To have a random starting point, call in the C file:
+ DrawTransitionMatrixFromPrior(smodel_est_ps);
+To mannually supply initial Q_k, (1) use the following 2-state 1-variable example in the input file datainp_markovswitching*.prn:
+ //== indxInitializeTransitionMatrix ==//
+ 1
+
+ //== Initial: Transition matrix ==//
+ 1.0 1.0
+ 0.0 0.0
+ (2) call DW's function in the C file:
+ ReadTransitionMatrices(fptr_markov, (char*)NULL, "Initial: ", smodel_ps); //Waggoner's function..
+ (3) for more complicated cases such as 2 independent state variables, we can have the following:
+ //== Initial: Transition matrix[1] ==//
+ 0.5 0.5
+ 0.5 0.5
+
+ //== Initial: Transition matrix[2] ==//
+ 0.9 0.2
+ 0.1 0.8
+
+
+
+//==================================
+// How to reset the probability of the initial state s_0
+//==================================
+At two places in switch.c, change the values (1: use ergodic distribution; 0: use 1/h for overall Q):
+ sv->UseErgodic=1; //in line 239
+ sv->UseErgodic=1; //in line 344
+
+
+//==================================
+// How to set the transition matrix for the 3rd state that is redundant to the 2nd state
+//==================================
+Say the estimated 2-state transition matrix is:
+
+ p11 p12
+ p21 p22
+
+Then the initial value of the 3-state transition matrix should be in the following form:
+
+ p11 p12 p12
+ p21 p22 p23
+ 0 0 p33
+
+where p23+p33 = p22.
+
+Matlab example about the ergodic properties of these transition matrices:
+x1 = [
+ 0.9 0.2
+ 0.1 0.8
+ ];
+e1=fn_ergodp(x)
+e1 =
+ 0.6667
+ 0.3333
+
+x2 = [
+ 0.9 0.2 0.2
+ 0.1 0.8 0.3
+ 0.0 0.0 0.5
+ ];
+e2=fn_ergodp(x2)
+e2 =
+ 0.6667
+ 0.3333
+ 0
+
+x2 = [
+ 0.9000 0.2000 0.2000
+ 0.1000 0.8000 0
+ 0 0 0.8000
+ ];
+e2=fn_ergodp(x2)
+e2 =
+ 0.6667
+ 0.3333
+ 0
+
+x2 = [
+ 0.9000 0.2000 0
+ 0.1000 0.8000 0.3000
+ 0 0 0.7000
+ ];
+e2=fn_ergodp(x2)
+e2 =
+ 0.6667
+ 0.3333
+ 0
+
+x2 = [
+ 0.9000 0.2000 0
+ 0.1000 0.8000 0
+ 0 0 1.0000
+ ];
+e2=fn_ergodp(x2)
+e2 =
+ 0
+ 0
+ 1
+
+e2=fn_ergodp(eye(4))
+e2 =
+ 0
+ 0
+ 0
+ 1
+
+
+
+//==================================
+// How to call base regimes
+//==================================
+int **Index = smodel_ps->sv->Index; //Regime-switching states.
+//--- Example 1 (two types of state variables) where s_t is an overall state variable combining the both state variables.
+st_d = st_dshocks = Index[s_t][0]; //[0]: 1st state variable.
+st_gain = Index[s_t][1]; //[1]: 2nd state variable.
+
+
+//==================================
+// How to write out transition matrices
+//==================================
+printf("Obj. func. value: %g\n", *fret_p);
+WriteTransitionMatrices(stdout, (char*)NULL, "DDDDebugging: ", SMODEL_PS); //Waggoner's function.
+fgetc(stdin); //fgetc(stdin) will automatically fflush(stdout). Use ctrl-c to stop the program.
+
+
+//==================================
+// DSGE output files such as simulations and impulse responses
+//==================================
+ Output format for ir_percentiles_TAG.prn reporting impulse responses with error bands
+-------------------------------------------------------------------------------------------------------
+Each block represents the responses corresponding to each percentile. In each block, the element in the position (k,i+j*nz) is the response of the ith state variable to the jth shock at horizon k. The variable nz is the total number of states. Shocks 0 through nepsilon-1 are fundamental and shocks nepsilon through nepsilon+nu-1 aremeasurement error. Horizon 0 is the contemporaneous response.
+
+This files contains the impulse responses of the unobservable variables. The code to produce the impulse responses of the observable variables has been commented out and so would be easy to produce.
+
+
+ Output format for simulation_TAG.prn reporting posterior draws of parameters
+-------------------------------------------------------------------------------------------------------
+Rows: posterior draws saved after the burning period and the thinning factor
+Column 1: log posterior (states integrated out)
+Column 2: log likelihood (states integrated out)
+Columns 3-length(Theta): free model specific parameters (Theta)
+Columns (last) free transition matrix parameters (q)
+
+The draws were generated with a burn-in period of 10,000 draws, this was starting from the posterior mode after the adaptive code had run to determine the jumping kernel. Example: two files contain 100,000 draws, but the thinning factor for the constant parameter case was 10 (1,000,000 total draws) while the thinning factor for the 2v case was 5 (500,000 total draws).
+
+
+ Output format for historical*_TAG.prn reporting historical decompositions.
+-------------------------------------------------------------------------------------------------------
+historical_TAG.csv: Historical decompositions of observables.
+ 1st colomn: dates. 0 -- beginning of the sample (include the lags), 1 -- second period, 2 -- third period, and so on.
+ 1st block (from the 2nd column on): concerns the 1st variable.
+ 1st column within this block: the data for the 1st variable.
+ 2nd column within this block: unconditional mean (with zero shocks) given the initial condition. The first lags rows are garbage and must be disposed.
+ 3rd - (2+nshocks)rd columns within this block: cumulative contributions of n shocks to this variable (deviations from the mean).
+ Note: 2nd column + 3rd - (2+nshocks)rd columns = 1st column.
+ 2nd block: concerns the 2nd variable.
+ 1st column within this block: ....
+ and so on.
+
+historicalstates_TAG.cvs: Historical decompositions of state variables (not observed).
+ 1st colomn: dates. 0 -- beginning of the sample (include the lags), 1 -- second period, 2 -- third period, and so on.
+ 1st block (from the 2nd column on): concerns the 1st variable.
+ 1st column within this block: smoothed value (mean value conditional on all the data up to T).
+ 2nd column within this block: unconditional mean (given the initial condition, NOT the data up to T) with zero shocks. The first lags rows are garbage and must be disposed.
+ 3rd - (2+nshocks)rd columns within this block: cumulative contributions of n shocks to this variable (deviations from the mean).
+ Note: 2nd column + 3rd - (2+nshocks)rd columns = 1st column.
+ 2nd block: concerns the 2nd variable.
+ 1st column within this block: ....
+ and so on.
+
+
+
+
+
+
+//-------------------------
+// How to use switch.c
+//-------------------------
+For call of the old functions, uncomment the following functions in switch.c
+
+TParameters* CreateParameters()
+
+
+
+
+
+
+
+Tempary fix
+in switch.h and switch.c
+ //=== Control variables ===
+ int UseErgodic; //1: use ergodic; 0: use 1/h for Q (not individual Q_k).
+
+===================================================================================================
+
+
+sv->UseErgodic=1; //in switch.c.
+
+//------------------------------------------------------------
+// Functions required by CreateParameters() in switch.c.
+//------------------------------------------------------------
+int NumberOfFreeModelSpecificParameters(struct TStateModel_tag *smodel_ps)
+{
+ struct TStvp7d_tag *tvp7d_ps = (struct TStvp7d_tag *)smodel_ps->p->p;
+ return (tvp7d_ps->gphi_dv->n);
+}
+//---
+void ConvertFreeParameters2ModelSpecificParameters(struct TStateModel_tag *smodel_ps, PRECISION *xopt_pd)
+{
+ //xopt_pd: the (temporary) vector of free model parameters for optimization or minimization.
+ struct TStvp7d_tag *tvp7d_ps = (struct TStvp7d_tag *)smodel_ps->p->p;
+ memcpy(tvp7d_ps->gphi_dv->v, xopt_pd, tvp7d_ps->gphi_dv->n);
+ $$$$$$ This gphi_dv must also be syncronized with interpretable individual parameters, which I forgot to do. ??????????
+ $$$$$$ Convertphi2indpars(tvp7d_ps);
+}
+//---
+void ConvertModelSpecificParameters2FreeParameters(struct TStateModel_tag *smodel_ps, PRECISION *xopt_pd)
+{
+ //xopt_pd: the (temporary) vector of free model parameters for optimization or minimization.
+ struct TStvp7d_tag *tvp7d_ps = (struct TStvp7d_tag *)smodel_ps->p->p;
+ $$$$$$$$$$$$$ Convertindpars2phi(tvp7d_ps);
+ memcpy(xopt_pd, tvp7d_ps->gphi_dv->v, tvp7d_ps->gphi_dv->n);
+}
+
+
+
+
+ConvertFreeParametersToTheta(smodel_ps, xphi_sdv.v); //Waggoner's function.
+ConvertFreeParametersToQ(smodel_ps, xqs_sdv.v); //Waggoner's function.
+smodel_ps->ValidForwardRecursion = 0; //Reset so recursion can do through.
+logvalue = LogLikelihoodGivenParameters(smodel_ps); //log(P(Y_T|theta, Q)). Waggoner's function.
+
diff --git a/tzDocumentation/testjunk.prn b/tzDocumentation/testjunk.prn
new file mode 100755
index 0000000000000000000000000000000000000000..2132a3bd5b15d9464a7fe881afedf4afcd4f0c4e
--- /dev/null
+++ b/tzDocumentation/testjunk.prn
@@ -0,0 +1,49 @@
+Example:
+
+swzmnhstage1 -fi MHM_input.dat -ft sz5v_2vby2m
+
+Command line:
+
+ Attempt to set up model from command line. Command line options are the
+ following
+
+ -di <directory>
+ If this argument exists, then all input files are in specified directory.
+
+ -do <directory>
+ If this argument exists, then all output files are in specified directory.
+
+ -ft <filename tag>
+ If this argument exists, then the following is attempted:
+
+ 1) specification file name: mhm_final_<tag>.dat
+ mhm arguments file name: mhm_final_<tag>.dat
+
+ 2) specification file name: mhm_intermediate_<tag>.dat
+ mhm arguments file name: mhm_intermediate_<tag>.dat
+
+ 3) specification file name: est_final_<tag>.dat
+ mhm arguments file name: -fi <filename>
+
+ -fi <filename>
+ If this argument exists, then additional mhm arguments are read from the
+ input file with the given filename.
+
+ -fs <filename>
+ If this argument exists, then the specification file name is <filename>.
+ The argument -ft takes precedence over -fs.
+
+ -fp <filename>
+ If this argument exists, then the posterior is read from <filename>. Must
+ be used in conjunction with the argument -fs. The default value is the
+ filename associated with the argument -fs.
+
+ -ph <header>
+ If this argument exists, then the header for the posterior file is
+ <header>. Must be used in conjuction with the arguments -fp or -fs. The
+ default value is "Posterior mode: ".
+
+ If no command line options are given, then attemps to use a default input file
+ with the name "default.ini". Returns one valid pointer to a TStateModel upon
+ success and null upon failure.
+