Merge branch 'test_speed' of git.dynare.org:JohannesPfeifer/dynare

Ref. !2344

matlab/estimation/online/online_auxiliary_filter.m

@@ -31,6 +31,7 @@ function online_auxiliary_filter(xparam1, dataset_, options_, M_, estim_params_,
 
 % Set seed for randn().
 options_ = set_dynare_seed_local_options(options_,'default');
+options_.verbosity=0; %particularly suppress warning messages during k_order_pert within the loop
 pruning = options_.particle.pruning;
 variance_update = true;
 online_opt = options_.posterior_sampler_options.current_options;
@@ -78,10 +79,10 @@ xparam = zeros(number_of_parameters,number_of_particles);
 Prior = dprior(bayestopt_, options_.prior_trunc);
 for i=1:number_of_particles
     info = 12042009;
-    while info
+    while info(1)
         candidate = Prior.draw();
         [info] = solve_model_for_online_filter(false, candidate, dataset_, options_, M_, estim_params_, bayestopt_, bounds, oo_.dr , oo_.steady_state, oo_.exo_steady_state, oo_.exo_det_steady_state);
-        if ~info
+        if ~info(1)
             xparam(:,i) = candidate(:);
         end
     end

tests/estimation/univariate/bayesian.mod

@@ -47,7 +47,7 @@ s2df = 1;
 // form expression for the joint posterior marginal distribution of β, a Gibbs
 // sampling algorithm is used (the prior for β and the inverse of σ² are independent).
 
-gibbslength = 1000000; // Set the number of iterations in Gibbs 
+gibbslength = 300000; // Set the number of iterations in Gibbs 
 burnin = 10000;        // Set the number of iterations to be discarded (try to remove the effects of the initial condition).
 steps = 10;            // Do not record all iterations (try to remove the dependence between the draws).
 

tests/estimation/univariate/bayesian_param_names.mod

@@ -50,14 +50,14 @@ s2df = 1;
 // form expression for the joint posterior marginal distribution of β, a Gibbs
 // sampling algorithm is used (the prior for β and the inverse of σ² are independent).
 
-gibbslength = 1000000; // Set the number of iterations in Gibbs 
+gibbslength = 100000; // Set the number of iterations in Gibbs 
 burnin = 10000;        // Set the number of iterations to be discarded (try to remove the effects of the initial condition).
 steps = 10;            // Do not record all iterations (try to remove the dependence between the draws).
 
 ds = olsgibbs(ds, 'eqols', beta0, V0, s2priormean, s2df, gibbslength, burnin, steps, {'eqols', 'eqols_olsgibbs_fit'}, 'olsgibbs_eq',{'b2'; 'b3'});
 
 // Since we use a diffuse prior for β, the posterior mean of β should be close to the OLS estimate.
-if max(abs(oo_.ols.ols_eq.beta-oo_.olsgibbs.olsgibbs_eq.posterior.mean.beta))>.1
+if max(abs(oo_.ols.ols_eq.beta-oo_.olsgibbs.olsgibbs_eq.posterior.mean.beta))>.01
    error('Something is wrong in the Gibbs sampling routine (univariate model)')
 end
 

tests/observation_trends_and_prefiltering/MCMC/Trend_loglin_no_prefilt_first_obs_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_loglin_no_prefilt_first_obs_MC_Exp_AR1_trend_
     mode_compute=4,silent_optimizer,first_obs=1000,loglinear,smoother,forecast=100,prefilter=0,
     mcmc_jumping_covariance='Trend_loglin_no_prefilt_first_obs_MC_MCMC_jump_covar',
     filtered_vars, filter_step_ahead = [1,2,4],
-    mh_nblocks=1,mh_jscale=0.3,no_posterior_kernel_density) P_obs Y_obs junk2;
+    mh_nblocks=1,mh_jscale=0.3,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
 load('Trend_loglin_no_prefilt_first_obs_MC_Exp_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 
 

tests/observation_trends_and_prefiltering/MCMC/Trend_loglin_prefilt_first_obs_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_loglin_prefilt_first_obs_MC_Exp_AR1_trend_dat
     mode_compute=4,silent_optimizer,first_obs=1000,loglinear,smoother,forecast=100,prefilter=1,
     mcmc_jumping_covariance='Trend_loglin_prefilt_first_obs_MC_MCMC_jump_covar_prefilter',
     filtered_vars, filter_step_ahead = [1,2,4],
-    mh_nblocks=1,mh_jscale=1e-4,no_posterior_kernel_density) P_obs Y_obs junk2;
+    mh_nblocks=1,mh_jscale=1e-4,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
     
 load('Trend_loglin_prefilt_first_obs_MC_Exp_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 

tests/observation_trends_and_prefiltering/MCMC/Trend_loglinear_no_prefilter_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_loglinear_no_prefilter_MC_Exp_AR1_trend_data_
     mode_compute=4,silent_optimizer,first_obs=1,loglinear,diffuse_filter,smoother,forecast=100,prefilter=0,
     mcmc_jumping_covariance='Trend_loglinear_no_prefilter_MC_MCMC_jump_covar',
     filtered_vars, filter_step_ahead = [1,2,4],
-    mh_nblocks=1,mh_jscale=0.3) P_obs Y_obs junk2;
+    mh_nblocks=1,mh_jscale=0.3,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
 load('Trend_loglinear_no_prefilter_MC_Exp_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 
 

tests/observation_trends_and_prefiltering/MCMC/Trend_loglinear_prefilter_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_loglinear_prefilter_MC_Exp_AR1_trend_data_wit
     mode_compute=4,silent_optimizer,first_obs=1,loglinear,smoother,forecast=100,prefilter=1,
     mcmc_jumping_covariance='Trend_loglinear_prefilter_MC_MCMC_jump_covar_prefilter',
     filtered_vars, filter_step_ahead = [1,2,4],
-    mh_nblocks=1,mh_jscale=1e-4) P_obs Y_obs junk2;
+    mh_nblocks=1,mh_jscale=1e-4,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
     
 load('Trend_loglinear_prefilter_MC_Exp_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 

tests/observation_trends_and_prefiltering/MCMC/Trend_no_prefilter_MC.mod

@@ -6,7 +6,7 @@ generate_trend_stationary_AR1(M_.fname);
 estimation(order=1,datafile='Trend_no_prefilter_MC_AR1_trend_data_with_constant',mh_replic=400,silent_optimizer,
             mode_compute=4,first_obs=1,smoother,mh_nblocks=1,mh_jscale=0.3,
             filtered_vars, filter_step_ahead = [1,2,4],
-            mcmc_jumping_covariance='Trend_no_prefilter_MC_MCMC_jump_covar',forecast=100,prefilter=0) P_obs Y_obs junk2;
+            mcmc_jumping_covariance='Trend_no_prefilter_MC_MCMC_jump_covar',forecast=100,prefilter=0,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
             
 load('Trend_no_prefilter_MC_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 

tests/observation_trends_and_prefiltering/MCMC/Trend_no_prefilter_first_obs_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_no_prefilter_first_obs_MC_AR1_trend_data_with
         mh_replic=400,mode_compute=4,silent_optimizer,first_obs=1000,smoother,forecast=100,prefilter=0,
         mcmc_jumping_covariance='Trend_no_prefilter_first_obs_MC_MCMC_jump_covar',
         filtered_vars, filter_step_ahead = [1,2,4],
-        mh_nblocks=1,mh_jscale=0.3,no_posterior_kernel_density) P_obs Y_obs junk2;
+        mh_nblocks=1,mh_jscale=0.3,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
 
 load('Trend_no_prefilter_first_obs_MC_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 

tests/observation_trends_and_prefiltering/MCMC/Trend_prefilter_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_prefilter_MC_AR1_trend_data_with_constant',mh
     first_obs=1,smoother,prefilter=1,
     mh_nblocks=1,mh_jscale=1e-4,
     filtered_vars, filter_step_ahead = [1,2,4],
-    mcmc_jumping_covariance='Trend_prefilter_MC_MCMC_jump_covar_prefilter',forecast=100) P_obs Y_obs junk2;
+    mcmc_jumping_covariance='Trend_prefilter_MC_MCMC_jump_covar_prefilter',forecast=100,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
 
 load('Trend_prefilter_MC_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 

tests/observation_trends_and_prefiltering/MCMC/Trend_prefilter_first_obs_MC.mod

@@ -7,7 +7,7 @@ estimation(order=1,datafile='Trend_prefilter_first_obs_MC_AR1_trend_data_with_co
         first_obs=1000,smoother,prefilter=1,
         mh_nblocks=1,mh_jscale=1e-4,
         filtered_vars, filter_step_ahead = [1,2,4],
-        mcmc_jumping_covariance='Trend_prefilter_first_obs_MC_MCMC_jump_covar_prefilter',forecast=100,no_posterior_kernel_density) P_obs Y_obs junk2;
+        mcmc_jumping_covariance='Trend_prefilter_first_obs_MC_MCMC_jump_covar_prefilter',forecast=100,no_posterior_kernel_density,nograph,sub_draws=100) P_obs Y_obs junk2;
 
 load('Trend_prefilter_first_obs_MC_AR1_trend_data_with_constant');
 @#include "../Trend_load_data_common.inc" 

tests/particle/dsge_base2.mod

@@ -105,7 +105,7 @@ varobs y l i ;
 %datatomfile('mysample')
 %return;
 
-data(file='./mysample.m',first_obs=801Y,nobs=200); %no measurement errors added in the simulated data
+data(file='./mysample.m',first_obs=801Y,nobs=50); %no measurement errors added in the simulated data
 
 @#if LINEAR_KALMAN
 	estimation(nograph,order=1,mode_compute=8,silent_optimizer,mh_replic=0,additional_optimizer_steps=[8 4],mode_check);
@@ -155,7 +155,7 @@ estimation(order=3,nograph,filter_algorithm=gf,proposal_approximation=montecarlo
 %  estimation(order=3,nograph,number_of_particles=10000,mode_compute=11,mh_replic=0,particle_filter_options=('liu_west_delta',0.9));
   estimation(order=1,posterior_sampling_method='online',posterior_sampler_options=('particles',1000));
   estimation(order=2,posterior_sampling_method='online',posterior_sampler_options=('particles',1000));
-  estimation(order=3,posterior_sampling_method='online',filter_algorithm=nlkf,proposal_approximation=montecarlo,number_of_particles=500,posterior_sampler_options=('particles',500));
+  estimation(order=3,posterior_sampling_method='online',filter_algorithm=nlkf,proposal_approximation=montecarlo,number_of_particles=100,posterior_sampler_options=('particles',100));
 @#endif
 
 @#if MCMC