function x0 = stab_map_(OutputDirectoryName,opt_gsa) % % function x0 = stab_map_(OutputDirectoryName) % % Mapping of stability regions in the prior ranges applying % Monte Carlo filtering techniques. % % INPUTS (from opt_gsa structure) % Nsam = MC sample size % fload = 0 to run new MC; 1 to load prevoiusly generated analysis % alpha2 = significance level for bivariate sensitivity analysis % [abs(corrcoef) > alpha2] % prepSA = 1: save transition matrices for mapping reduced form % = 0: no transition matrix saved (default) % pprior = 1: sample from prior ranges (default): sample saved in % _prior.mat file % = 0: sample from posterior ranges: sample saved in % _mc.mat file % OUTPUT: % x0: one parameter vector for which the model is stable. % % GRAPHS % 1) Pdf's of marginal distributions under the stability (dotted % lines) and unstability (solid lines) regions % 2) Cumulative distributions of: % - stable subset (dotted lines) % - unacceptable subset (solid lines) % 3) Bivariate plots of significant correlation patterns % ( abs(corrcoef) > alpha2) under the stable and unacceptable subsets % % USES qmc_sequence, stab_map_1, stab_map_2 % % Written by Marco Ratto % Joint Research Centre, The European Commission, % marco.ratto@ec.europa.eu % Copyright (C) 2012-2016 European Commission % Copyright (C) 2012-2018 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 . %global bayestopt_ estim_params_ dr_ options_ ys_ fname_ global bayestopt_ estim_params_ options_ oo_ M_ % opt_gsa=options_.opt_gsa; Nsam = opt_gsa.Nsam; fload = opt_gsa.load_stab; alpha2 = opt_gsa.alpha2_stab; pvalue_ks = opt_gsa.pvalue_ks; pvalue_corr = opt_gsa.pvalue_corr; prepSA = (opt_gsa.redform | opt_gsa.identification); pprior = opt_gsa.pprior; neighborhood_width = opt_gsa.neighborhood_width; ilptau = opt_gsa.ilptau; nliv = opt_gsa.morris_nliv; ntra = opt_gsa.morris_ntra; dr_ = oo_.dr; %if isfield(dr_,'ghx'), ys_ = oo_.dr.ys; nspred = M_.nspred; %size(dr_.ghx,2); nboth = M_.nboth; nfwrd = M_.nfwrd; %end fname_ = M_.fname; np = estim_params_.np; nshock = estim_params_.nvx; nshock = nshock + estim_params_.nvn; nshock = nshock + estim_params_.ncx; nshock = nshock + estim_params_.ncn; lpmat0=zeros(Nsam,0); xparam1=[]; pshape = bayestopt_.pshape(nshock+1:end); p1 = bayestopt_.p1(nshock+1:end); p2 = bayestopt_.p2(nshock+1:end); p3 = bayestopt_.p3(nshock+1:end); p4 = bayestopt_.p4(nshock+1:end); [~,~,~,lb,ub,~] = set_prior(estim_params_,M_,options_); %Prepare bounds if ~isempty(bayestopt_) && any(bayestopt_.pshape > 0) % Set prior bounds bounds = prior_bounds(bayestopt_, options_.prior_trunc); bounds.lb = max(bounds.lb,lb); bounds.ub = min(bounds.ub,ub); else % estimated parameters but no declared priors % No priors are declared so Dynare will estimate the model by % maximum likelihood with inequality constraints for the parameters. bounds.lb = lb; bounds.ub = ub; if opt_gsa.prior_range==0 warning('GSA:: When using ML, sampling from the prior is not possible. Setting prior_range=1') opt_gsa.prior_range=1; end end if nargin==0 OutputDirectoryName=''; end options_mcf.pvalue_ks = pvalue_ks; options_mcf.pvalue_corr = pvalue_corr; options_mcf.alpha2 = alpha2; name=cell(np,1); name_tex=cell(np,1); for jj=1:np if options_.TeX [param_name_temp, param_name_tex_temp]= get_the_name(nshock+jj,options_.TeX,M_,estim_params_,options_); name_tex{jj,1} = strrep(param_name_tex_temp,'$',''); name{jj,1} = param_name_temp; else param_name_temp = get_the_name(nshock+jj,options_.TeX,M_,estim_params_,options_); name{jj,1} = param_name_temp; end end if options_.TeX options_mcf.param_names_tex = name_tex; end options_mcf.param_names = name; options_mcf.fname_ = fname_; options_mcf.OutputDirectoryName = OutputDirectoryName; options_mcf.xparam1 = []; opt=options_; options_.periods=0; options_.nomoments=1; options_.irf=0; options_.noprint=1; if fload==0 % if prepSA % T=zeros(size(dr_.ghx,1),size(dr_.ghx,2)+size(dr_.ghu,2),Nsam/2); % end if isfield(dr_,'ghx') egg=zeros(length(dr_.eigval),Nsam); end yys=zeros(length(dr_.ys),Nsam); if opt_gsa.morris == 1 [lpmat, OutFact] = Sampling_Function_2(nliv, np+nshock, ntra, ones(np+nshock, 1), zeros(np+nshock,1), []); lpmat = lpmat.*(nliv-1)/nliv+1/nliv/2; Nsam=size(lpmat,1); lpmat0 = lpmat(:,1:nshock); lpmat = lpmat(:,nshock+1:end); % elseif opt_gsa.morris==3, % lpmat = prep_ide(Nsam,np,5); % Nsam=size(lpmat,1); else if np<52 && ilptau>0 [lpmat] = qmc_sequence(np, int64(1), 0, Nsam)'; if np>30 || ilptau==2 % scrambled lptau for j=1:np lpmat(:,j)=lpmat(randperm(Nsam),j); end end else %ilptau==0 [lpmat] = NaN(Nsam,np); for j=1:np lpmat(:,j) = randperm(Nsam)'./(Nsam+1); %latin hypercube end end end % try dummy=prior_draw_gsa(1); %initialize persistent variables % catch % if pprior, % if opt_gsa.prior_range==0; % error('Some unknown prior is specified or ML estimation,: use prior_range=1 option!!'); % end % end % % end if pprior for j=1:nshock if opt_gsa.morris~=1 lpmat0(:,j) = randperm(Nsam)'./(Nsam+1); %latin hypercube end if opt_gsa.prior_range lpmat0(:,j)=lpmat0(:,j).*(bounds.ub(j)-bounds.lb(j))+bounds.lb(j); end end if opt_gsa.prior_range % if opt_gsa.identification, % deltx=min(0.001, 1/Nsam/2); % for j=1:np, % xdelt(:,:,j)=prior_draw_gsa(0,[lpmat0 lpmat]+deltx); % end % end for j=1:np lpmat(:,j)=lpmat(:,j).*(bounds.ub(j+nshock)-bounds.lb(j+nshock))+bounds.lb(j+nshock); end else xx=prior_draw_gsa(0,[lpmat0 lpmat]); % if opt_gsa.identification, % deltx=min(0.001, 1/Nsam/2); % ldum=[lpmat0 lpmat]; % ldum = prior_draw_gsa(0,ldum+deltx); % for j=1:nshock+np, % xdelt(:,:,j)=xx; % xdelt(:,j,j)=ldum(:,j); % end % clear ldum % end lpmat0=xx(:,1:nshock); lpmat=xx(:,nshock+1:end); clear xx; end else % for j=1:nshock, % xparam1(j) = oo_.posterior_mode.shocks_std.(bayestopt_.name{j}); % sd(j) = oo_.posterior_std.shocks_std.(bayestopt_.name{j}); % lpmat0(:,j) = randperm(Nsam)'./(Nsam+1); %latin hypercube % lb = max(bayestopt_.lb(j), xparam1(j)-2*sd(j)); % ub1=xparam1(j)+(xparam1(j) - lb); % define symmetric range around the mode! % ub = min(bayestopt_.ub(j),ub1); % if ub30 & np<52 % lpmat(:,j) = lpmat(randperm(Nsam),j).*(ub-lb)+lb; % else % lpmat(:,j) = lpmat(:,j).*(ub-lb)+lb; % end % end %load([fname_,'_mode']) if neighborhood_width>0 && isempty(options_.mode_file) xparam1 = get_all_parameters(estim_params_,M_); else eval(['load ' options_.mode_file '.mat;']); end if neighborhood_width>0 for j=1:nshock if opt_gsa.morris ~= 1 lpmat0(:,j) = randperm(Nsam)'./(Nsam+1); %latin hypercube end ub=min([bounds.ub(j) xparam1(j)*(1+neighborhood_width)]); lb=max([bounds.lb(j) xparam1(j)*(1-neighborhood_width)]); lpmat0(:,j)=lpmat0(:,j).*(ub-lb)+lb; end for j=1:np ub=xparam1(j+nshock)*(1+sign(xparam1(j+nshock))*neighborhood_width); lb=xparam1(j+nshock)*(1-sign(xparam1(j+nshock))*neighborhood_width); if bounds.ub(j+nshock)>=xparam1(j) && bounds.lb(j)<=xparam1(j+nshock) ub=min([bounds.ub(j+nshock) ub]); lb=max([bounds.lb(j+nshock) lb]); else fprintf('\nstab_map_:: the calibrated value of param %s for neighborhood_width sampling is outside prior bounds.\nWe allow violation of bounds for this parameter, but if this was not done on purpose, please change calibration before running neighborhood_width sampling\n', bayestopt_.name{j+nshock}) end lpmat(:,j)=lpmat(:,j).*(ub-lb)+lb; end else d = chol(inv(hh)); lp=randn(Nsam*2,nshock+np)*d+kron(ones(Nsam*2,1),xparam1'); lnprior=zeros(1,Nsam*2); for j=1:Nsam*2 lnprior(j) = any(lp(j,:)'<=bounds.lb | lp(j,:)'>=bounds.ub); end ireal=[1:2*Nsam]; ireal=ireal(find(lnprior==0)); lp=lp(ireal,:); Nsam=min(Nsam, length(ireal)); lpmat0=lp(1:Nsam,1:nshock); lpmat=lp(1:Nsam,nshock+1:end); clear lp lnprior ireal; end end % h = dyn_waitbar(0,'Please wait...'); istable=[1:Nsam]; jstab=0; iunstable=[1:Nsam]; iindeterm=zeros(1,Nsam); iwrong=zeros(1,Nsam); inorestriction=zeros(1,Nsam); irestriction=zeros(1,Nsam); infox=zeros(Nsam,1); for j=1:Nsam M_ = set_all_parameters([lpmat0(j,:) lpmat(j,:)]',estim_params_,M_); %try stoch_simul([]); try if ~ isempty(options_.endogenous_prior_restrictions.moment) [Tt,Rr,SteadyState,info,M_,options_,oo_] = dynare_resolve(M_,options_,oo_); else [Tt,Rr,SteadyState,info,M_,options_,oo_] = dynare_resolve(M_,options_,oo_,'restrict'); end infox(j,1)=info(1); if infox(j,1)==0 && ~exist('T','var') dr_=oo_.dr; if prepSA try T=zeros(size(dr_.ghx,1),size(dr_.ghx,2)+size(dr_.ghu,2),Nsam); catch ME = lasterror(); if strcmp('MATLAB:nomem',ME.identifier) prepSA=0; disp('The model is too large for storing state space matrices ...') disp('for mapping reduced form or for identification') end T=[]; end else T=[]; end egg=zeros(length(dr_.eigval),Nsam); end if infox(j,1) % disp('no solution'), if isfield(oo_.dr,'ghx') oo_.dr=rmfield(oo_.dr,'ghx'); end if (infox(j,1)<3 || infox(j,1)>5) && isfield(oo_.dr,'eigval') oo_.dr=rmfield(oo_.dr,'eigval'); end end catch ME if isfield(oo_.dr,'eigval') oo_.dr=rmfield(oo_.dr,'eigval'); end if isfield(oo_.dr,'ghx') oo_.dr=rmfield(oo_.dr,'ghx'); end disp('No solution could be found') end dr_ = oo_.dr; if isfield(dr_,'ghx') egg(:,j) = sort(dr_.eigval); if prepSA jstab=jstab+1; T(:,:,jstab) = [dr_.ghx dr_.ghu]; % [A,B] = ghx2transition(squeeze(T(:,:,jstab)), ... % bayestopt_.restrict_var_list, ... % bayestopt_.restrict_columns, ... % bayestopt_.restrict_aux); end if ~exist('nspred','var') nspred = dr_.nspred; %size(dr_.ghx,2); nboth = dr_.nboth; nfwrd = dr_.nfwrd; end info=endogenous_prior_restrictions(Tt,Rr,M_,options_,oo_); infox(j,1)=info(1); if info(1) inorestriction(j)=j; else iunstable(j)=0; irestriction(j)=j; end else istable(j)=0; if isfield(dr_,'eigval') egg(:,j) = sort(dr_.eigval); if exist('nspred','var') if any(isnan(egg(1:nspred,j))) iwrong(j)=j; else if (nboth || nfwrd) && abs(egg(nspred+1,j))<=options_.qz_criterium iindeterm(j)=j; end end end else if exist('egg','var') egg(:,j)=ones(size(egg,1),1).*NaN; end iwrong(j)=j; end end ys_=real(dr_.ys); yys(:,j) = ys_; ys_=yys(:,1); dyn_waitbar(j/Nsam,h,['MC iteration ',int2str(j),'/',int2str(Nsam)]) end dyn_waitbar_close(h); if prepSA && jstab T=T(:,:,1:jstab); else T=[]; end istable=istable(find(istable)); % stable params ignoring restrictions irestriction=irestriction(find(irestriction)); % stable params & restrictions OK inorestriction=inorestriction(find(inorestriction)); % stable params violating restrictions iunstable=iunstable(find(iunstable)); % violation of BK & restrictions & solution could not be found (whatever goes wrong) iindeterm=iindeterm(find(iindeterm)); % indeterminacy iwrong=iwrong(find(iwrong)); % dynare could not find solution ixun=iunstable(find(~ismember(iunstable,[iindeterm,iwrong,inorestriction]))); % explosive roots % % map stable samples % istable=[1:Nsam]; % for j=1:Nsam, % if any(isnan(egg(1:nspred,j))) % istable(j)=0; % else % if abs(egg(nspred,j))>=options_.qz_criterium; %(1-(options_.qz_criterium-1)); %1-1.e-5; % istable(j)=0; % %elseif (dr_.nboth | dr_.nfwrd) & abs(egg(nspred+1,j))<=options_.qz_criterium; %1+1.e-5; % elseif (nboth | nfwrd) & abs(egg(nspred+1,j))<=options_.qz_criterium; %1+1.e-5; % istable(j)=0; % end % end % end % istable=istable(find(istable)); % stable params % % % map unstable samples % iunstable=[1:Nsam]; % for j=1:Nsam, % %if abs(egg(dr_.npred+1,j))>1+1.e-5 & abs(egg(dr_.npred,j))<1-1.e-5; % %if (dr_.nboth | dr_.nfwrd), % if ~any(isnan(egg(1:5,j))) % if (nboth | nfwrd), % if abs(egg(nspred+1,j))>options_.qz_criterium & abs(egg(nspred,j))0 || length(iwrong)>0 fprintf(['%4.1f%% of the prior support gives unique saddle-path solution.\n'],length(istable)/Nsam*100) fprintf(['%4.1f%% of the prior support gives explosive dynamics.\n'],(length(ixun) )/Nsam*100) if ~isempty(iindeterm) fprintf(['%4.1f%% of the prior support gives indeterminacy.\n'],length(iindeterm)/Nsam*100) end inorestriction = istable(find(~ismember(istable,irestriction))); % violation of prior restrictions if ~isempty(iwrong) || ~isempty(inorestriction) skipline() if any(infox==49) fprintf(['%4.1f%% of the prior support violates prior restrictions.\n'],(length(inorestriction) )/Nsam*100) end if ~isempty(iwrong) skipline() disp(['For ',num2str(length(iwrong)/Nsam*100,'%4.1f'),'% of the prior support dynare could not find a solution.']) skipline() end if any(infox==1) disp([' For ',num2str(length(find(infox==1))/Nsam*100,'%4.1f'),'% The model doesn''t determine the current variables uniquely.']) end if any(infox==2) disp([' For ',num2str(length(find(infox==2))/Nsam*100,'%4.1f'),'% MJDGGES returned an error code.']) end if any(infox==6) disp([' For ',num2str(length(find(infox==6))/Nsam*100,'%4.1f'),'% The jacobian evaluated at the deterministic steady state is complex.']) end if any(infox==19) disp([' For ',num2str(length(find(infox==19))/Nsam*100,'%4.1f'),'% The steadystate routine has thrown an exception (inconsistent deep parameters).']) end if any(infox==20) disp([' For ',num2str(length(find(infox==20))/Nsam*100,'%4.1f'),'% Cannot find the steady state.']) end if any(infox==21) disp([' For ',num2str(length(find(infox==21))/Nsam*100,'%4.1f'),'% The steady state is complex.']) end if any(infox==22) disp([' For ',num2str(length(find(infox==22))/Nsam*100,'%4.1f'),'% The steady has NaNs.']) end if any(infox==23) disp([' For ',num2str(length(find(infox==23))/Nsam*100,'%4.1f'),'% M_.params has been updated in the steadystate routine and has complex valued scalars.']) end if any(infox==24) disp([' For ',num2str(length(find(infox==24))/Nsam*100,'%4.1f'),'% M_.params has been updated in the steadystate routine and has some NaNs.']) end if any(infox==30) disp([' For ',num2str(length(find(infox==30))/Nsam*100,'%4.1f'),'% Ergodic variance can''t be computed.']) end end skipline() if length(iunstable)1 itot = [1:Nsam]; isolve = itot(find(~ismember(itot,iwrong))); % dynare could find a solution % Blanchard Kahn if neighborhood_width options_mcf.xparam1 = xparam1(nshock+1:end); end itmp = itot(find(~ismember(itot,istable))); options_mcf.amcf_name = asname; options_mcf.amcf_title = atitle; options_mcf.beha_title = 'unique Stable Saddle-Path'; options_mcf.nobeha_title = 'NO unique Stable Saddle-Path'; options_mcf.title = 'unique solution'; mcf_analysis(lpmat, istable, itmp, options_mcf, options_); if ~isempty(iindeterm) itmp = isolve(find(~ismember(isolve,iindeterm))); options_mcf.amcf_name = aindname; options_mcf.amcf_title = aindtitle; options_mcf.beha_title = 'NO indeterminacy'; options_mcf.nobeha_title = 'indeterminacy'; options_mcf.title = 'indeterminacy'; mcf_analysis(lpmat, itmp, iindeterm, options_mcf, options_); end if ~isempty(ixun) itmp = isolve(find(~ismember(isolve,ixun))); options_mcf.amcf_name = aunstname; options_mcf.amcf_title = aunsttitle; options_mcf.beha_title = 'NO explosive solution'; options_mcf.nobeha_title = 'explosive solution'; options_mcf.title = 'instability'; mcf_analysis(lpmat, itmp, ixun, options_mcf, options_); end inorestriction = istable(find(~ismember(istable,irestriction))); % violation of prior restrictions iwrong = iwrong(find(~ismember(iwrong,inorestriction))); % what went wrong beyond prior restrictions if ~isempty(iwrong) itmp = itot(find(~ismember(itot,iwrong))); options_mcf.amcf_name = awrongname; options_mcf.amcf_title = awrongtitle; options_mcf.beha_title = 'NO inability to find a solution'; options_mcf.nobeha_title = 'inability to find a solution'; options_mcf.title = 'inability to find a solution'; mcf_analysis(lpmat, itmp, iwrong, options_mcf, options_); end if ~isempty(irestriction) if neighborhood_width options_mcf.xparam1 = xparam1; end np=size(bayestopt_.name,1); name=cell(np,1); name_tex=cell(np,1); for jj=1:np if options_.TeX [param_name_temp, param_name_tex_temp]= get_the_name(jj,options_.TeX,M_,estim_params_,options_); name_tex{jj,1} = strrep(param_name_tex_temp,'$',''); name{jj,1} = param_name_temp; else param_name_temp = get_the_name(jj,options_.TeX,M_,estim_params_,options_); name{jj,1} = param_name_temp; end end if options_.TeX options_mcf.param_names_tex = name_tex; end options_mcf.param_names = name; options_mcf.amcf_name = acalibname; options_mcf.amcf_title = acalibtitle; options_mcf.beha_title = 'prior IRF/moment calibration'; options_mcf.nobeha_title = 'NO prior IRF/moment calibration'; options_mcf.title = 'prior restrictions'; mcf_analysis([lpmat0 lpmat], irestriction, inorestriction, options_mcf, options_); iok = irestriction(1); x0 = [lpmat0(iok,:)'; lpmat(iok,:)']; else iok = istable(1); x0=0.5.*(bounds.ub(1:nshock)-bounds.lb(1:nshock))+bounds.lb(1:nshock); x0 = [x0; lpmat(iok,:)']; end M_ = set_all_parameters(x0,estim_params_,M_); [oo_.dr,info,M_,options_,oo_] = resol(0,M_,options_,oo_); % stoch_simul([]); else disp('All parameter values in the specified ranges are not acceptable!') x0=[]; end else disp('All parameter values in the specified ranges give unique saddle-path solution,') disp('and match prior IRF/moment restriction(s) if any!') x0=0.5.*(bounds.ub(1:nshock)-bounds.lb(1:nshock))+bounds.lb(1:nshock); x0 = [x0; lpmat(istable(1),:)']; end xparam1=x0; save prior_ok.mat xparam1; options_.periods=opt.periods; if isfield(opt,'nomoments') options_.nomoments=opt.nomoments; end options_.irf=opt.irf; options_.noprint=opt.noprint;