dynare/matlab/prior_sampler.m

function results = prior_sampler(drsave,M_,bayestopt_,options_,oo_)
% This function builds a (big) prior sample.
%
% INPUTS
%   drsave      [integer]    Scalar. If equal to 1, then dr structure is saved with each prior draw.     
%   M_          [structure]  Model description.
%   bayestopt_  [structure]  Prior distribution description.  
%   options_    [structure]  Global options of Dynare.
%    
% OUTPUTS:
%   results     [structure]  Various statistics. 
%
% SPECIAL REQUIREMENTS
%   none

% Copyright (C) 2009 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/>.
    
    % Initialization.
    prior_draw(1,bayestopt_);
    PriorDirectoryName = CheckPath('prior/draws');
    work = ~drsave;
    iteration = 1;
    loop_indx = 0;
    file_indx = [];
    count_bk_indeterminacy = 0;
    count_bk_unstability = 0;
    count_bk_singularity = 0;
    count_static_var_def = 0;
    count_no_steadystate = 0;
    count_steadystate_file_exit = 0;
    count_dll_problem = 0;
    count_complex_jacobian = 0;
    count_complex_steadystate = 0;
    count_unknown_problem = 0;
    NumberOfSimulations = options_.prior_mc;
    NumberOfParameters = length(bayestopt_.p1);
    NumberOfEndogenousVariables = size(M_.endo_names,1);
    NumberOfElementsPerFile = ceil(options_.MaxNumberOfBytes/NumberOfParameters/NumberOfEndogenousVariables/8) ;
    
    if NumberOfSimulations <= NumberOfElementsPerFile
        TableOfInformations = [ 1 ,  NumberOfSimulations , 1] ;
    else
        NumberOfFiles = fix(NumberOfSimulations/NumberOfElementsPerFile) ;
        NumberOfElementsInTheLastFile = NumberOfSimulations - NumberOfElementsPerFile*(NumberOfFiles-1) ;
        if ~isint(NumberOfSimulations/NumberOfElementsPerFile)
            NumberOfFiles = NumberOfFiles + 1 ;
        end
        TableOfInformations = NaN(NumberOfFiles,3);
        TableOfInformations(:,1) = transpose(1:NumberOfFiles) ;
        TableOfInformations(1:NumberOfFiles-1,2) = NumberOfElementsPerFile*ones(NumberOfFiles-1,1) ;
        TableOfInformations(NumberOfFiles,2) = NumberOfElementsInTheLastFile ;
        TableOfInformations(1,3) = 1;
        TableOfInformations(2:end,3) = cumsum(TableOfInformations(2:end,2))+1;
    end
    
    pdraws = cell(TableOfInformations(1,2),drsave+1) ;
    sampled_prior_expectation = zeros(NumberOfParameters,1);
    sampled_prior_covariance  = zeros(NumberOfParameters,NumberOfParameters);
    
    % Simulations.
    while iteration <= NumberOfSimulations
        loop_indx = loop_indx+1;
        file_indx = find(TableOfInformations(:,3)==iteration);
        if ~isempty(file_indx) && file_indx>1
            save([ PriorDirectoryName '/prior_draws' int2str(file_indx-1) '.mat' ],'pdraws');
            pdraws = cell(TableOfInformations(file_indx,2),drsave+1);
        end
        params = prior_draw();
        set_all_parameters(params);
        [dr,INFO] = resol(oo_.steady_state,work);
        switch INFO(1)
          case 0
            pdraws(iteration,1) = {params};
            if drsave
                pdraws(iteration,2) = {dr};
            end
            iteration = iteration+1;
            [sampled_prior_expectation,sampled_prior_covariance] = ...
                recursive_prior_moments(sampled_prior_expectation,sampled_prior_covariance,params,iteration) ;
          case 1
            count_static_undefined = count_static_undefined + 1;
          case 2
            count_dll_problem = count_dll_problem + 1;
          case 3
            count_bk_unstability = count_bk_unstability + 1 ;
          case 4
            count_bk_indeterminacy = count_bk_indeterminacy + 1 ;
          case 5
            count_bk_singularity = count_bk_singularity + 1 ;
          case 20
            count_no_steadystate = count_no_steadystate + 1 ;
          case 19
            count_steadystate_file_exit = count_steadystate_file_exit + 1 ;
          case 6
            count_complex_jacobian = count_complex_jacobian + 1 ;
          case 21
            count_complex_steadystate = count_complex_steadystate + 1 ;
          otherwise
            count_unknown_problem = count_unknown_problem + 1 ;
        end
    end
    
    % Save last prior_draw*.mat file
    save([ PriorDirectoryName '/prior_draws' int2str(TableOfInformations(end,1)) '.mat' ],'pdraws');
    
    % Get informations about BK conditions and other things...
    results.bk.indeterminacy_share = count_bk_indeterminacy/loop_indx;
    results.bk.unstability_share = count_bk_unstability/loop_indx;
    results.bk.singularity_share = count_bk_singularity/loop_indx;
    results.dll.problem_share = count_dll_problem/loop_indx;
    results.ss.problem_share = count_no_steadystate/loop_indx;
    results.ss.complex_share = count_complex_steadystate/loop_indx;
    results.ass.problem_share = count_steadystate_file_exit/loop_indx;
    results.jacobian.problem_share = count_complex_jacobian/loop_indx;
    results.garbage_share = ...
        results.bk.indeterminacy_share + ...
        results.bk.unstability_share + ...
        results.bk.singularity_share + ...
        results.dll.problem_share + ...
        results.ss.problem_share + ...
        results.ass.problem_share + ...
        results.jacobian.problem_share + ...
        count_unknown_problem/loop_indx ;
    results.prior.mean = sampled_prior_expectation;
    results.prior.variance = sampled_prior_covariance;
    results.prior.mass = 1-results.garbage_share;
    
    
function [mu,sigma] = recursive_prior_moments(m0,s0,newobs,iter)
%  Recursive estimation of order one and two moments (expectation and
%  covariance matrix). newobs should be a row vector. I do not use the
%  function recursive_moments here, because this function is to be used when
%  newobs is a 2D array.
    m1 = m0 + (newobs'-m0)/iter;
    qq = m1*m1';
    s1 = s0 + ( (newobs'*newobs-qq-s0) + (iter-1)*(m0*m0'-qq') )/iter;
    mu = m1;
    sigma = s1;