stepan
/
dynare
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

v4 draw_prior_density.m: 

* factorized code for PDFs
* multiplied by 10 the truncation for inv gamma type 2 (so as to mimic the truncation for inv gamma type 1)
* various cosmetic changes


git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@1984 ac1d8469-bf42-47a9-8791-bf33cf982152
time-shift
sebastien 2008-08-05 13:31:33 +00:00
parent b61c9e6a69
commit 19e9bb3bd1
1 changed files with 27 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

62
matlab/draw_prior_density.m
View File

@ -1,6 +1,6 @@
function [x,f,abscissa,dens,binf,bsup] = draw_prior_density(indx);
% function [x,f,abscissa,dens,binf,bsup] = draw_prior_density(indx)
% plots prior density
% Computes values of prior density at many points (before plotting)
%
% INPUTS
% indx: parameter number
@ -10,8 +10,8 @@ function [x,f,abscissa,dens,binf,bsup] = draw_prior_density(indx);
% f: subset of 'dens' such as the density is less than 10
% abscissa: abscissa
% dens: density
% binf: lower bound of the truncated prior
% bsup: upper bound of the truncated prior
% binf: first element of x
% bsup: last element of x
%
% SPECIAL REQUIREMENTS
% none
@ -42,11 +42,12 @@ p2 = bayestopt_.p2;
p3 = bayestopt_.p3;
p4 = bayestopt_.p4;
truncprior = 10^(-3);
truncprior = 1e-3;
steps = 200;
switch pshape(indx)
case 1 % Beta prior
density = inline('((bb-x).^(b-1)).*(x-aa).^(a-1)./(beta(a,b)*(bb-aa)^(a+b-1))','x','a','b','aa','bb');
density = @(x,a,b,aa,bb) betapdf((x-aa)/(bb-aa), a, b)/(bb-aa);
mu = (p1(indx)-p3(indx))/(p4(indx)-p3(indx));
stdd = p2(indx)/(p4(indx)-p3(indx));
a = (1-mu)*mu^2/stdd^2 - mu;
@ -55,59 +56,50 @@ switch pshape(indx)
bb = p4(indx);
infbound = betainv(truncprior,a,b)*(bb-aa)+aa;
supbound = betainv(1-truncprior,a,b)*(bb-aa)+aa;
stepsize = (supbound-infbound)/200;
stepsize = (supbound-infbound)/steps;
abscissa = infbound:stepsize:supbound;
dens = density(abscissa,a,b,aa,bb);
case 2 % Generalized Gamma prior
density = @(x,a,b,c) gampdf(x-c,a,b);
mu = p1(indx)-p3(indx);
b = p2(indx)^2/mu;
a = mu/b;
infbound = gaminv(truncprior,a,b);
supbound = gaminv(1-truncprior,a,b);
stepsize = (supbound-infbound)/200;
c = p3(indx);
infbound = gaminv(truncprior,a,b)+c;
supbound = gaminv(1-truncprior,a,b)+c;
stepsize = (supbound-infbound)/steps;
abscissa = infbound:stepsize:supbound;
dens = exp(lpdfgam(abscissa,a,b));
abscissa = abscissa + p3(indx);
dens = density(abscissa,a,b,c);
case 3 % Gaussian prior
density = inline('inv(sqrt(2*pi)*b)*exp(-0.5*((x-a)/b).^2)','x','a','b');
a = p1(indx);
b = p2(indx);
infbound = norminv(truncprior,a,b);
supbound = norminv(1-truncprior,a,b);
stepsize = (supbound-infbound)/200;
stepsize = (supbound-infbound)/steps;
abscissa = infbound:stepsize:supbound;
dens = density(abscissa,a,b);
dens = normpdf(abscissa,a,b);
case 4 % Inverse-gamma of type 1 prior
density = inline('2*inv(gamma(nu/2))*(x.^(-nu-1))*((s/2)^(nu/2)).*exp(-s./(2*x.^2))','x','s','nu');
nu = p2(indx);
s = p1(indx);
a = nu/2;
b = 2/s;
infbound = 1/sqrt(gaminv(1-10*truncprior,a,b));
supbound = 1/sqrt(gaminv(10*truncprior,a,b));
stepsize = (supbound-infbound)/200;
infbound = 1/sqrt(gaminv(1-10*truncprior, nu/2, 2/s));
supbound = 1/sqrt(gaminv(10*truncprior, nu/2, 2/s));
stepsize = (supbound-infbound)/steps;
abscissa = infbound:stepsize:supbound;
dens = density(abscissa,s,nu);
dens = exp(lpdfig1(abscissa,s,nu));
case 5 % Uniform prior
density = inline('(x.^0)/(b-a)','x','a','b');
a = p1(indx);
b = p2(indx);
infbound = a;
supbound = b;
stepsize = (supbound-infbound)/200;
infbound = p1(indx);
supbound = p2(indx);
stepsize = (supbound-infbound)/steps;
abscissa = infbound:stepsize:supbound;
dens = density(abscissa,a,b);
dens = ones(1, steps) / (supbound-infbound);
case 6 % Inverse-gamma of type 2 prior
density = inline('inv(gamma(nu/2))*(x.^(-.5*(nu+2)))*((s/2)^(nu/2)).*exp(-s./(2*x))','x','s','nu');
nu = p2(indx);
s = p1(indx);
a = nu/2;
b = 2/s;
infbound = 1/(gaminv(1-truncprior,a,b));
supbound = 1/(gaminv(truncprior,a,b));
stepsize = (supbound-infbound)/200;
infbound = 1/(gaminv(1-10*truncprior, nu/2, 2/s));
supbound = 1/(gaminv(10*truncprior, nu/2, 2/s));
stepsize = (supbound-infbound)/steps;
abscissa = infbound:stepsize:supbound;
dens = density(abscissa,s,nu);
dens = exp(lpdfig2(abscissa,s,nu));
otherwise
error(sprintf('draw_prior_density: unknown distribution shape (index %d, type %d)', indx, pshape(indx)));
end