Select Git revision
bvar_forecast.m
-
adjemian authored
v4.1: Posterior forecasts plots of bvar à la Sims report the median instead of the mean (because the median is less affected by the posterior draws such that the eigenvalues of the companion matrix are outside the unit circle). git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@2057 ac1d8469-bf42-47a9-8791-bf33cf982152
adjemian authoredv4.1: Posterior forecasts plots of bvar à la Sims report the median instead of the mean (because the median is less affected by the posterior draws such that the eigenvalues of the companion matrix are outside the unit circle). git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@2057 ac1d8469-bf42-47a9-8791-bf33cf982152
draw_prior_density.m 4.58 KiB
function [x,f,abscissa,dens,binf,bsup] = draw_prior_density(indx,bayestopt_);
% Computes values of prior densities at many points (before plotting)
%
% INPUTS
% indx [integer] Parameter number.
% bayestopt_ [structure] Describes the prior beliefs.
%
% OUTPUTS
% x [double] Row vector, subset of 'abscissa' such as the density is less than 10
% f [double] Row vector, subset of 'dens' such as the density is less than 10
% abscissa [double] Row vector, abscissa
% dens [double] Row vector, density
% binf: [double] Scalar, first element of x
% bsup: [double] Scalar, last element of x
% Copyright (C) 2004-2011 Dynare Team
%
% This file is part of Dynare.
%
% Dynare is free software: you can redistribute it and/or modify
% it under the terms of the GNU General Public License as published by
% the Free Software Foundation, either version 3 of the License, or
% (at your option) any later version.
%
% Dynare is distributed in the hope that it will be useful,
% but WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
% GNU General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
pshape = bayestopt_.pshape;
p3 = bayestopt_.p3;
p4 = bayestopt_.p4;
p6 = bayestopt_.p6;
p7 = bayestopt_.p7;
truncprior = 1e-3;
steps = 200;
switch pshape(indx)
case 1% Beta prior
density = @(x,a,b,aa,bb) betapdf((x-aa)/(bb-aa), a, b)/(bb-aa);
infbound = betainv(truncprior,p6(indx),p7(indx))*(p4(indx)-p3(indx))+p3(indx);
supbound = betainv(1-truncprior,p6(indx),p7(indx))*(p4(indx)-p3(indx))+p3(indx);
abscissa = linspace(infbound,supbound,steps);
dens = density(abscissa,p6(indx),p7(indx),p3(indx),p4(indx));
case 2% Generalized Gamma prior
density = @(x,a,b,c) gampdf(x-c,a,b);
try
infbound = gaminv(truncprior,p6(indx),p7(indx))+p3(indx);
supbound = gaminv(1-truncprior,p6(indx),p7(indx))+p3(indx);
catch
% Workaround for ticket #161
if exist('OCTAVE_VERSION')
error(['Due to a bug in Octave, you must choose other values for mean and/or variance of your prior on ' bayestopt_.name{indx} ', or use another shape'])
else
rethrow(lasterror)
end
end
abscissa = linspace(infbound,supbound,steps);
dens = density(abscissa,p6(indx),p7(indx),p3(indx));
case 3% Gaussian prior
infbound = norminv(truncprior,p6(indx),p7(indx));
supbound = norminv(1-truncprior,p6(indx),p7(indx));
abscissa = linspace(infbound,supbound,steps);
dens = normpdf(abscissa,p6(indx),p7(indx));
case 4% Inverse-gamma of type 1 prior
try
infbound = 1/sqrt(gaminv(1-10*truncprior, p7(indx)/2, 2/p6(indx)))+p3(indx);
supbound = 1/sqrt(gaminv(10*truncprior, p7(indx)/2, 2/p6(indx)))+p3(indx);
catch
% Workaround for ticket #161
if exist('OCTAVE_VERSION')
error(['Due to a bug in Octave, you must choose other values for mean and/or variance of your prior on ' bayestopt_.name{indx} ', or use another shape'])
else
rethrow(lasterror)
end
end
abscissa = linspace(infbound,supbound,steps);
dens = exp(lpdfig1(abscissa-p3(indx),p6(indx),p7(indx)));
case 5% Uniform prior
infbound = p6(indx);
supbound = p7(indx);
abscissa = linspace(infbound,supbound,steps);
dens = ones(1, steps) / (supbound-infbound);
case 6% Inverse-gamma of type 2 prior
try
infbound = 1/(gaminv(1-10*truncprior, p7(indx)/2, 2/p6(indx)))+p3(indx);
supbound = 1/(gaminv(10*truncprior, p7(indx)/2, 2/p6(indx)))+p3(indx);
catch
% Workaround for ticket #161
if exist('OCTAVE_VERSION')
error(['Due to a bug in Octave, you must choose other values for mean and/or variance of your prior on ' bayestopt_.name{indx} ', or use another shape'])
else
rethrow(lasterror)
end
end
abscissa = linspace(infbound,supbound,steps);
dens = exp(lpdfig2(abscissa-p3(indx),p6(indx),p7(indx)));
otherwise
error(sprintf('draw_prior_density: unknown distribution shape (index %d, type %d)', indx, pshape(indx)));
end
if pshape(indx) ~= 5
[junk,k1] = max(dens);
if k1 == 1 || k1 == length(dens)
k = find(dens > 10);
dens(k) = NaN;
end
end
binf = abscissa(1);
bsup = abscissa(end);
x = abscissa;
f = dens;
f(find(x<bayestopt_.lb(indx)))=0;
f(find(x>bayestopt_.ub(indx)))=0;