dynare/matlab/+occbin/DSGE_smoother.m

406 lines
20 KiB
Matlab

function [alphahat,etahat,epsilonhat,ahat0,SteadyState,trend_coeff,aKK,T0,R0,P,PKK,decomp,Trend,state_uncertainty,M_,oo_,bayestopt_] = DSGE_smoother(xparam1,gend,Y,data_index,missing_value,M_,oo_,options_,bayestopt_,estim_params_,dataset_, dataset_info)
%function [alphahat,etahat,epsilonhat,ahat0,SteadyState,trend_coeff,aKK,T0,R0,P,PKK,decomp,Trend,state_uncertainty,M_,oo_,bayestopt_] = DSGE_smoother(xparam1,gend,Y,data_index,missing_value,M_,oo_,options_,bayestopt_,estim_params_,dataset_, dataset_info)
% Runs a DSGE smoother with occasionally binding constraints
%
% INPUTS
% - xparam1 [double] (p*1) vector of (estimated) parameters.
% - gend [integer] scalar specifying the number of observations
% - Y [double] (n*T) matrix of data.
% - data_index [cell] 1*smpl cell of column vectors of indices.
% - missing_value [boolean] 1 if missing values, 0 otherwise
% - M_ [structure] Matlab's structure describing the model (M_).
% - oo_ [structure] Matlab's structure containing the results (oo_).
% - options_ [structure] Matlab's structure describing the current options (options_).
% - bayestopt_ [structure] describing the priors
% - estim_params_ [structure] characterizing parameters to be estimated
% - dataset_ [structure] the dataset after required transformation
% - dataset_info [structure] Various informations about the dataset (descriptive statistics and missing observations)
%
% OUTPUTS
% - alphahat [double] (m*T) matrix, smoothed endogenous variables (a_{t|T}) (decision-rule order)
% - etahat [double] (r*T) matrix, smoothed structural shocks (r>=n is the number of shocks).
% - epsilonhat [double] (n*T) matrix, smoothed measurement errors.
% - ahat0 [double] (m*T) matrix, updated (endogenous) variables (a_{t|t}) (decision-rule order)
% - SteadyState [double] (m*1) vector specifying the steady state level of each endogenous variable (declaration order)
% - trend_coeff [double] (n*1) vector, parameters specifying the slope of the trend associated to each observed variable.
% - aKK [double] (K,n,T+K) array, k (k=1,...,K) steps ahead
% filtered (endogenous) variables (decision-rule order)
% - T0 and R0 [double] Matrices defining the state equation (T is the (m*m) transition matrix).
% - P: (m*m*(T+1)) 3D array of one-step ahead forecast error variance
% matrices (decision-rule order)
% - PKK (K*m*m*(T+K)) 4D array of k-step ahead forecast error variance
% matrices (meaningless for periods 1:d) (decision-rule order)
% - decomp (K*m*r*(T+K)) 4D array of shock decomposition of k-step ahead
% filtered variables (decision-rule order)
% - Trend [double] (n*T) pure trend component; stored in options_.varobs order
% - state_uncertainty [double] (K,K,T) array, storing the uncertainty
% about the smoothed state (decision-rule order)
% - M_ [structure] decribing the model
% - oo_ [structure] storing the results
% - options_ [structure] describing the options
% - bayestopt_ [structure] describing the priors
% Copyright © 2021 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 <https://www.gnu.org/licenses/>.
smoother_field_list = {'SmoothedVariables', 'UpdatedVariables', 'SmoothedShocks'};
regime_history=[];
if options_.occbin.smoother.linear_smoother && nargin==12
%% linear smoother
options_.occbin.smoother.status=false;
[alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,T0,R0,P,PK,decomp,Trend,state_uncertainty,M_,oo_,bayestopt_] = DsgeSmoother(xparam1,gend,Y,data_index,missing_value,M_,oo_,options_,bayestopt_,estim_params_);
tmp_smoother=store_smoother_results(M_,oo_,options_,bayestopt_,dataset_,dataset_info,alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,P,PK,decomp,Trend,state_uncertainty);
for jf=1:length(smoother_field_list)
oo_.occbin.linear_smoother.(smoother_field_list{jf}) = tmp_smoother.(smoother_field_list{jf});
end
oo_.occbin.linear_smoother.alphahat=alphahat;
oo_.occbin.linear_smoother.etahat=etahat;
oo_.occbin.linear_smoother.epsilonhat=epsilonhat;
oo_.occbin.linear_smoother.ahat=ahat;
oo_.occbin.linear_smoother.SteadyState=SteadyState;
oo_.occbin.linear_smoother.trend_coeff=trend_coeff;
oo_.occbin.linear_smoother.aK=aK;
oo_.occbin.linear_smoother.T0=T0;
oo_.occbin.linear_smoother.R0=R0;
oo_.occbin.linear_smoother.decomp=decomp;
fprintf('\nOccbin: linear smoother done.\n')
options_.occbin.smoother.status=true;
end
% if init_mode
%% keep these commented lines for the moment, should some nasty problem in the future call back for this option
% shocks1 = etahat(:,2:end)';
% clear regime_history ;
%
% for ii = 1:gend-1
%
% fprintf('Period number %d out of %d \n',ii,gend-1)
% opts_simul.SHOCKS = shocks1(ii,:);
% opts_simul.endo_init = alphahat(oo_.dr.inv_order_var,ii);
% out = occbin.runsim_fn(opts_simul, M_, oo_, options_);
% regime_history(ii) = out.regime_history;
% end
%
% [TT,RR] = dynare_resolve(M_,options_,oo_);
% CC = zeros([size(RR,1),gend ]);
% else
opts_simul = options_.occbin.simul;
opts_simul.curb_retrench = options_.occbin.smoother.curb_retrench;
opts_simul.maxit = options_.occbin.smoother.maxit;
opts_simul.waitbar = options_.occbin.smoother.waitbar;
opts_simul.periods = options_.occbin.smoother.periods;
opts_simul.check_ahead_periods = options_.occbin.smoother.check_ahead_periods;
opts_simul.periodic_solution = options_.occbin.smoother.periodic_solution;
opts_simul.full_output = options_.occbin.smoother.full_output;
opts_simul.piecewise_only = options_.occbin.smoother.piecewise_only;
% init_mode = options_.occbin.smoother.init_mode; % 0 = standard; 1 = unconditional frcsts zero shocks+smoothed states in each period
% init_mode = 0;
occbin_options = struct();
occbin_options.first_period_occbin_update = options_.occbin.smoother.first_period_occbin_update;
occbin_options.opts_regime = opts_simul; % this builds the opts_simul options field needed by occbin.solver
occbin_options.opts_regime.binding_indicator = options_.occbin.likelihood.init_binding_indicator;
occbin_options.opts_regime.regime_history=options_.occbin.likelihood.init_regime_history;
[alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,T0,R0,P,PK,decomp,Trend,state_uncertainty,M_,oo_,bayestopt_] = DsgeSmoother(xparam1,gend,Y,data_index,missing_value,M_,oo_,options_,bayestopt_,estim_params_,occbin_options);% T1=TT;
oo_.occbin.smoother.realtime_regime_history = oo_.occbin.smoother.regime_history;
regime_history = oo_.occbin.smoother.regime_history;
opts_regime.regime_history = oo_.occbin.smoother.regime_history;
ahat0 = ahat;
aKK=aK;
PKK=PK;
clear aK PK;
occbin_options.first_period_occbin_update = inf;
opts_regime.binding_indicator=[];
regime_history0 = regime_history;
fprintf('Occbin smoother iteration 1.\n')
opts_simul.SHOCKS = [etahat(:,2:end)'; zeros(1,M_.exo_nbr)];
opts_simul.exo_pos = 1:M_.exo_nbr;
opts_simul.endo_init = alphahat(oo_.dr.inv_order_var,1);
opts_simul.init_regime=regime_history; % use realtime regime for guess, to avoid multiple solution issues!
options_.occbin.simul=opts_simul;
[~, out, ss] = occbin.solver(M_,oo_,options_);
regime_history = out.regime_history;
if options_.smoother_redux
occbin_options.opts_simul.restrict_state_space =1;
oo_.occbin.linear_smoother.T0=ss.T(oo_.dr.order_var,oo_.dr.order_var,1);
oo_.occbin.linear_smoother.R0=ss.R(oo_.dr.order_var,:,1);
end
TT = ss.T(oo_.dr.order_var,oo_.dr.order_var,:);
RR = ss.R(oo_.dr.order_var,:,:);
CC = ss.C(oo_.dr.order_var,:);
TT = cat(3,TT(:,:,1),TT);
RR = cat(3,RR(:,:,1),RR);
CC = cat(2,CC(:,1),CC);
opts_regime.regime_history = regime_history;
opts_regime.binding_indicator = [];
[TT, RR, CC, regime_history] = occbin.check_regimes(TT, RR, CC, opts_regime, M_, oo_, options_);
is_changed = ~isequal(regime_history0,regime_history);
if isempty(regime_history0)
regime_history0 = regime_history;
end
iter=1;
is_periodic = 0;
is_changed_start = 0;
maxiter = options_.occbin.smoother.max_number_of_iterations;
occbin_smoother_fast = options_.occbin.smoother.fast;
occbin_smoother_debug=options_.occbin.smoother.debug;
sto_alphahat=alphahat;
sto_etahat={etahat};
sto_CC = CC;
sto_RR = RR;
sto_TT = TT;
for k=1:size(TT,3)
sto_eee(:,k) = eig(TT(:,:,k));
end
while is_changed && maxiter>iter && ~is_periodic
iter=iter+1;
fprintf('Occbin smoother iteration %u.\n', iter)
occbin_options.opts_regime.regime_history=regime_history;
[alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,T0,R0,P,PK,decomp,Trend,state_uncertainty,M_,oo_,bayestopt_] = DsgeSmoother(xparam1,gend,Y,data_index,missing_value,M_,oo_,options_,bayestopt_,estim_params_,occbin_options,TT,RR,CC);% T1=TT;
sto_etahat(iter)={etahat};
regime_history0(iter,:) = regime_history;
save('info1','regime_history0');
sto_CC = CC;
sto_RR = RR;
sto_TT = TT;
opts_simul.SHOCKS = [etahat(:,2:end)'; zeros(1,M_.exo_nbr)];
opts_simul.endo_init = alphahat(oo_.dr.inv_order_var,1);
options_.occbin.simul=opts_simul;
[~, out, ss] = occbin.solver(M_,oo_,options_);
regime_history = out.regime_history;
TT = ss.T(oo_.dr.order_var,oo_.dr.order_var,:);
RR = ss.R(oo_.dr.order_var,:,:);
CC = ss.C(oo_.dr.order_var,:);
TT = cat(3,TT(:,:,1),TT);
RR = cat(3,RR(:,:,1),RR);
CC = cat(2,CC(:,1),CC);
opts_regime.regime_history = regime_history;
[TT, RR, CC, regime_history] = occbin.check_regimes(TT, RR, CC, opts_regime, M_, oo_, options_);
is_changed = ~isequal(regime_history0(iter,:),regime_history);
if M_.occbin.constraint_nbr==2
for k=1:size(regime_history0,2)
isdiff_regime(k,1) = ~isequal(regime_history0(end,k).regime1,regime_history(k).regime1);
isdiff_start(k,1) = ~isequal(regime_history0(end,k).regimestart1,regime_history(k).regimestart1);
isdiff_(k,1) = isdiff_regime(k,1) || isdiff_start(k,1);
isdiff_regime(k,2) = ~isequal(regime_history0(end,k).regime2,regime_history(k).regime2);
isdiff_start(k,2) = ~isequal(regime_history0(end,k).regimestart2,regime_history(k).regimestart2);
isdiff_(k,2) = isdiff_regime(k,2) || isdiff_start(k,2);
end
is_changed_regime = ~isempty(find(isdiff_regime(:,1))) || ~isempty(find(isdiff_regime(:,2)));
is_changed_start = ~isempty(find(isdiff_start(:,1))) || ~isempty(find(isdiff_start(:,2)));
else
for k=1:size(regime_history0,2)
isdiff_regime(k,1) = ~isequal(regime_history0(end,k).regime,regime_history(k).regime);
isdiff_start(k,1) = ~isequal(regime_history0(end,k).regimestart,regime_history(k).regimestart);
isdiff_(k,1) = isdiff_regime(k,1) || isdiff_start(k,1);
end
is_changed_regime = ~isempty(find(isdiff_regime(:,1)));
is_changed_start = ~isempty(find(isdiff_start(:,1)));
end
if occbin_smoother_fast
is_changed = is_changed_regime;
end
if iter>1
for kiter=1:iter-1
is_tmp = isequal(regime_history0(kiter,:),regime_history);
if is_tmp
break
end
end
is_periodic = (is_changed && is_tmp);
end
if is_changed
for k=1:size(TT,3)
eee(:,k) = eig(TT(:,:,k));
end
err_eig(iter-1) = max(max(abs(sort(eee)-sort(sto_eee))));
err_alphahat(iter-1) = max(max(max(abs(alphahat-sto_alphahat))));
err_etahat(iter-1) = max(max(max(abs(etahat-sto_etahat{iter-1}))));
err_CC(iter-1) = max(max(max(abs(CC-sto_CC))));
err_RR(iter-1) = max(max(max(abs(RR-sto_RR))));
err_TT(iter-1) = max(max(max(abs(TT-sto_TT))));
end
if occbin_smoother_debug
regime_ = cell(0);
regime_new = regime_;
start_ = regime_;
start_new = regime_;
if M_.occbin.constraint_nbr==2
indx_init_1 = find(isdiff_(:,1));
if ~isempty(indx_init_1)
qq={regime_history0(end,indx_init_1).regime1}';
for j=1:length(qq), regime_(j,1) = {int2str(qq{j})}; end
qq={regime_history0(end,indx_init_1).regimestart1}';
for j=1:length(qq), start_(j,1) = {int2str(qq{j})}; end
qq={regime_history(indx_init_1).regime1}';
for j=1:length(qq), regime_new(j,1) = {int2str(qq{j})}; end
qq={regime_history(indx_init_1).regimestart1}';
for j=1:length(qq), start_new(j,1) = {int2str(qq{j})}; end
disp('Time points where regime 1 differs')
if ~isoctave
disp(table(indx_init_1, regime_, start_, regime_new, start_new))
else % The table() function is not implemented in Octave, print something more or less equivalent (though much less readable)
disp(vertcat({'indx_init_1', 'regime_', 'start_', 'regime_new', 'start_new'}, ...
horzcat(num2cell(indx_init_1), regime_, start_, regime_new, start_new)))
end
end
indx_init_2 = find(isdiff_(:,2));
if ~isempty(indx_init_2)
regime_ = cell(0);
regime_new = regime_;
start_ = regime_;
start_new = regime_;
qq={regime_history0(end,indx_init_2).regime2}';
for j=1:length(qq), regime_(j,1) = {int2str(qq{j})}; end
qq={regime_history0(end,indx_init_2).regimestart2}';
for j=1:length(qq), start_(j,1) = {int2str(qq{j})}; end
qq={regime_history(indx_init_2).regime2}';
for j=1:length(qq), regime_new(j,1) = {int2str(qq{j})}; end
qq={regime_history(indx_init_2).regimestart2}';
for j=1:length(qq), start_new(j,1) = {int2str(qq{j})}; end
disp('Time points where regime 2 differs ')
if ~isoctave
disp(table(indx_init_2, regime_, start_, regime_new, start_new))
else % The table() function is not implemented in Octave, print something more or less equivalent (though much less readable)
disp(vertcat({'indx_init_2', 'regime_', 'start_', 'regime_new', 'start_new'}, ...
horzcat(num2cell(indx_init_2), regime_, start_, regime_new, start_new)))
end
end
else
indx_init_1 = find(isdiff_(:,1));
if ~isempty(indx_init_1)
qq={regime_history0(end,indx_init_1).regime}';
for j=1:length(qq), regime_(j,1) = {int2str(qq{j})}; end
qq={regime_history0(end,indx_init_1).regimestart}';
for j=1:length(qq), start_(j,1) = {int2str(qq{j})}; end
qq={regime_history(indx_init_1).regime}';
for j=1:length(qq), regime_new(j,1) = {int2str(qq{j})}; end
qq={regime_history(indx_init_1).regimestart}';
for j=1:length(qq), start_new(j,1) = {int2str(qq{j})}; end
disp('Time points where regime differs')
if ~isoctave
disp(table(indx_init_1, regime_, start_, regime_new, start_new))
else % The table() function is not implemented in Octave, print something more or less equivalent (though much less readable)
disp(vertcat({'indx_init_1', 'regime_', 'start_', 'regime_new', 'start_new'}, ...
horzcat(num2cell(indx_init_1), regime_, start_, regime_new, start_new)))
end
end
end
end
if is_changed
sto_alphahat=alphahat;
sto_eee = eee;
end
end
regime_history0(max(iter+1,1),:) = regime_history;
oo_.occbin.smoother.regime_history=regime_history0(end,:);
oo_.occbin.smoother.regime_history_iter=regime_history0;
save('info1','regime_history0')
if (maxiter==iter && is_changed) || is_periodic
disp(['Occbin smoother did not converge.'])
if is_periodic
disp(['Occbin smoother algo loops between two solutions.'])
end
else
disp(['Occbin smoother converged.'])
if occbin_smoother_fast && is_changed_start
disp('WARNING: fast algo is used, regime(s) duration(s) was not forced to converge')
end
end
if (~is_changed || occbin_smoother_debug) && nargin==12
if is_changed
CC = sto_CC;
RR = sto_RR;
TT = sto_TT;
oo_.occbin.smoother.regime_history=regime_history0(end-1,:);
end
tmp_smoother=store_smoother_results(M_,oo_,options_,bayestopt_,dataset_,dataset_info,alphahat,etahat,epsilonhat,ahat0,SteadyState,trend_coeff,aKK,P,PKK,decomp,Trend,state_uncertainty);
for jf=1:length(smoother_field_list)
oo_.occbin.smoother.(smoother_field_list{jf}) = tmp_smoother.(smoother_field_list{jf});
end
oo_.occbin.smoother.alphahat=alphahat;
oo_.occbin.smoother.etahat=etahat;
oo_.occbin.smoother.epsilonhat=epsilonhat;
oo_.occbin.smoother.ahat=ahat0;
oo_.occbin.smoother.SteadyState=SteadyState;
oo_.occbin.smoother.trend_coeff=trend_coeff;
oo_.occbin.smoother.aK=aKK;
oo_.occbin.smoother.T0=TT;
oo_.occbin.smoother.R0=RR;
oo_.occbin.smoother.C0=CC;
if options_.occbin.smoother.plot
GraphDirectoryName = CheckPath('graphs',M_.fname);
j1=0;
ifig=0;
for j=1:M_.exo_nbr
if M_.Sigma_e(j,j)
j1=j1+1;
if mod(j1,9)==1
hfig = dyn_figure(options_.nodisplay,'name','Occbin smoothed shocks');
ifig=ifig+1;
isub=0;
end
isub=isub+1;
subplot(3,3,isub)
if options_.occbin.smoother.linear_smoother
plot(oo_.occbin.linear_smoother.etahat(j,:)','linewidth',2)
hold on,
end
plot(oo_.occbin.smoother.etahat(j,:)','r--','linewidth',2)
hold on, plot([0 options_.nobs],[0 0],'k--')
set(gca,'xlim',[0 options_.nobs])
title(deblank(M_.exo_names(j,:)),'interpreter','none')
if mod(j1,9)==0
if options_.occbin.smoother.linear_smoother
annotation('textbox', [0.1,0,0.35,0.05],'String', 'Linear','Color','Blue','horizontalalignment','center','interpreter','none');
end
annotation('textbox', [0.55,0,0.35,0.05],'String', 'Piecewise','Color','Red','horizontalalignment','center','interpreter','none');
dyn_saveas(gcf,[GraphDirectoryName filesep M_.fname,'_smoothedshocks_occbin',int2str(ifig)],options_.nodisplay,options_.graph_format);
end
end
if mod(j1,9)~=0 && j==M_.exo_nbr
annotation('textbox', [0.1,0,0.35,0.05],'String', 'Linear','Color','Blue','horizontalalignment','center','interpreter','none');
annotation('textbox', [0.55,0,0.35,0.05],'String', 'Piecewise','Color','Red','horizontalalignment','center','interpreter','none');
dyn_saveas(hfig,[GraphDirectoryName filesep M_.fname,'_smoothedshocks_occbin',int2str(ifig)],options_.nodisplay,options_.graph_format);
end
end
end
end