dynare/matlab/smoother2histval.m

263 lines
9.3 KiB
Matlab

function smoother2histval(opts)
% This function takes values from oo_.SmoothedVariables (and possibly
% oo_.SmoothedShocks) and copies them into M_.histval.
%
% Optional fields in 'opts' structure:
% infile: An optional *_results MAT file created by Dynare.
% If present, oo_.Smoothed{Variables,Shocks} are read from
% there. Otherwise, they are read from the global workspace.
% invars: An optional char or cell array listing variables to read in
% oo_.SmoothedVariables. If absent, all the endogenous
% variables present in oo_.SmoothedVariables are used.
% period: An optional period number to use as the starting point
% for subsequent simulations. It should be between 1 and
% the number of observations that were used to produce the
% smoothed values. If absent, the last observation is used.
% outfile: An optional MAT file in which to save the histval structure.
% If absent, the output will be written in M_.endo_histval
% outvars: An optional char or cell array listing variables to be written in
% outfile or M_.endo_histval. This cell must be of same
% length than invars, and there is a mapping between the input
% variable at the i-th position in invars, and the output
% variable at the i-th position in outvars. If absent, then
% taken as equal to invars.
%
% The function also uses the value of option_.parameter_set
% Copyright © 2014-2021 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/>.
global M_ options_ oo_
if ~isfield(opts, 'infile')
if ~isfield(oo_, 'SmoothedVariables')
error('Could not find smoothed variables; did you set the "smoother" option?')
end
smoothedvars = oo_.SmoothedVariables;
smoothedshocks = oo_.SmoothedShocks;
else
S = load(opts.infile);
if ~isfield(S, 'oo_') || ~isfield(S.oo_, 'SmoothedVariables')
error('Could not find smoothed variables in file; is this a Dynare results file, and did you set the "smoother" option when producing it?')
end
smoothedvars = S.oo_.SmoothedVariables;
smoothedshocks = S.oo_.SmoothedShocks;
end
% Hack to determine if oo_.SmoothedVariables was computed after a Metropolis
tmp = fieldnames(smoothedvars);
if isstruct(smoothedvars.(tmp{1}))
post_metropolis = 1;
if ~ isstruct(smoothedvars.(tmp{end}))
% point and metropolis results are simultaneously present
post_metropolis = 2;
end
elseif isstruct(smoothedvars.(tmp{end}))
% point and metropolis results are simultaneously present
post_metropolis = 2;
else
post_metropolis = 0;
end
if post_metropolis
tmp = fieldnames(smoothedvars.Mean);
if length(tmp)~=M_.endo_nbr
warning(['You are using smoother2histval although smoothed values have not '...
'been computed for all endogenous and auxiliary variables.'...
'The value of these variables will be set to their steady state.'])
end
tmpexo = fieldnames(smoothedshocks.Mean);
else
tmp = fieldnames(smoothedvars);
tmpexo = fieldnames(smoothedshocks);
end
% If post-Metropolis, select the parameter set
if isempty(options_.parameter_set)
if post_metropolis
smoothedvars = smoothedvars.Mean;
smoothedshocks = smoothedshocks.Mean;
end
else
switch options_.parameter_set
case 'calibration'
if post_metropolis == 1
error('Option parameter_set=calibration is not consistent with computed smoothed values.')
end
case 'posterior_mode'
if post_metropolis == 1
error('Option parameter_set=posterior_mode is not consistent with computed smoothed values.')
end
case 'posterior_mean'
if ~post_metropolis
error('Option parameter_set=posterior_mean is not consistent with computed smoothed values.')
end
smoothedvars = smoothedvars.Mean;
smoothedshocks = smoothedshocks.Mean;
case 'posterior_median'
if ~post_metropolis
error('Option parameter_set=posterior_median is not consistent with computed smoothed values.')
end
smoothedvars = smoothedvars.Median;
smoothedshocks = smoothedshocks.Median;
otherwise
error([ 'Option parameter_set=' options_.parameter_set ' unsupported.' ])
end
end
% Determine number of periods
n = length(smoothedvars.(tmp{1}));
if n < M_.maximum_endo_lag
error('Not enough observations to create initial conditions')
end
if isfield(opts, 'invars')
invars = opts.invars;
if ischar(invars)
invars = cellstr(invars);
end
else
invars = [tmp; tmpexo];
end
if isfield(opts, 'period')
period = opts.period;
if period > n
error('The period that you indicated is beyond the data sample')
end
if period < M_.maximum_endo_lag
error('The period that you indicated is too small to construct initial conditions')
end
else
period = n;
end
if isfield(opts, 'outvars')
outvars = opts.outvars;
if ischar(outvars)
outvars = cellstr(outvars);
end
if length(invars) ~= length(outvars)
error('The number of input and output variables is not the same')
end
else
outvars = invars;
end
% Initialize outputs
if ~isfield(opts, 'outfile')
% Output to M_.endo_histval
M_.endo_histval = repmat(oo_.steady_state, 1, M_.maximum_lag);
else
% Output to a file
data = zeros(M_.maximum_endo_lag, length(invars));
for i=1:length(outvars)
j = strmatch(outvars{i}, M_.endo_names, 'exact');
if ~isempty(j)
data(:,i)=oo_.steady_state(j);
end
end
o = dseries();
end
% Handle all endogenous variables to be copied
for i = 1:length(invars)
if ~isempty(strmatch(invars{i}, M_.endo_names, 'exact'))
if oo_.Smoother.loglinear
s = exp(smoothedvars.(invars{i}));
else
s = smoothedvars.(invars{i});
end
elseif ~isempty(strmatch(invars{i}, M_.exo_names, 'exact'))
s = smoothedshocks.(invars{i});
else
error('smoother2histval: unknown input variable')
end
v = s((period-M_.maximum_lag+1):period);
if ~isfield(opts, 'outfile')
j_endo = strmatch(outvars{i}, M_.endo_names, 'exact');
if ~isempty(j_endo)
M_.endo_histval(j_endo, :) = v;
end
j_exo = strmatch(outvars{i}, M_.exo_names, 'exact');
if ~isempty(j_exo)
M_.exo_histval(j_exo, :) = v;
end
if isempty(j_endo) && isempty(j_exo)
error(['smoother2histval: output variable ' outvars{i} ' does not exist.'])
end
else
data(:, i) = v';
end
end
if isfield(opts, 'outfile')
o = dseries(data, '1Y', invars);
end
% $$$ % Handle auxiliary variables for lags (both on endogenous and exogenous)
% $$$ for i = 1:length(M_.aux_vars)
% $$$ if ~ ismember(M_.endo_names{M_.aux_vars(i).endo_index},invars)
% $$$ if M_.aux_vars(i).type ~= 1 && M_.aux_vars(i).type ~= 3
% $$$ continue
% $$$ end
% $$$ if M_.aux_vars(i).type == 1
% $$$ % Endogenous
% $$$ orig_var = M_.endo_names{M_.aux_vars(i).orig_index};
% $$$ else
% $$$ % Exogenous
% $$$ orig_var = M_.exo_names{M_.aux_vars(i).orig_index};
% $$$ end
% $$$ [m, k] = ismember(orig_var, outvars);
% $$$ if m
% $$$ if ~isempty(strmatch(invars{k}, M_.endo_names))
% $$$ s = smoothedvars.(invars{k});
% $$$ else
% $$$ s = smoothedshocks.(invars{k});
% $$$ end
% $$$ l = M_.aux_vars(i).orig_lead_lag;
% $$$ if period-M_.maximum_endo_lag+1+l < 1
% $$$ error('The period that you indicated is too small to construct initial conditions')
% $$$ end
% $$$ j = M_.aux_vars(i).endo_index;
% $$$ v = s((period-M_.maximum_endo_lag+1+l):(period+l)); %+steady_state(j);
% $$$ if ~isfield(opts, 'outfile')
% $$$ M_.endo_histval(j, :) = v;
% $$$ else
% $$$ % When saving to a file, x(-2) is in the variable called "x_l2"
% $$$ lead_lag = num2str(l);
% $$$ lead_lag = regexprep(lead_lag, '-', 'l');
% $$$ o.([ orig_var '_' lead_lag ]) = v;
% $$$ end
% $$$ end
% $$$ end
% $$$ end
% Finalize output
if isfield(opts, 'outfile')
[dir, fname, ext] = fileparts(opts.outfile);
if ~strcmp(ext,'.mat') && ~isempty(ext)
error(['smoother2hisvtval: if outfile has an extension, it must ' ...
'be .mat'])
end
o.save([dir fname]);
end
end