var: fix AR, EC matrices
Houtan Bastani
parent
90105af62b
commit
6384c9636c
|
|
@ -0,0 +1,243 @@
|
|||
function get_ar_ec_matrices(var_model_name)
|
||||
%function get_ar_ec_matrices(var_model_name)
|
||||
%
|
||||
% Returns the autoregressive and error correction matrices associated with the
|
||||
% VAR specified by var_model_name. Output is stored in cellarray
|
||||
% oo_.var.(var_model_name).ar, with oo_.var.(var_model_name).ar{i} being the
|
||||
% AR matrix at time t-i (same holds for error correction matrices with ec
|
||||
% replacing ar). Each AR (EC) matrix is stored with rows organized by the
|
||||
% ordering of the equation tags found in M_.var.(var_model_name).eqtags and
|
||||
% columns organized consistently.
|
||||
%
|
||||
% INPUTS
|
||||
%
|
||||
% var_model_name [string] the name of the VAR model
|
||||
%
|
||||
% OUTPUTS
|
||||
%
|
||||
% NONE
|
||||
|
||||
% Copyright (C) 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 <http://www.gnu.org/licenses/>.
|
||||
|
||||
global M_ oo_
|
||||
|
||||
%% Check inputs and initialize output
|
||||
assert(nargin == 1, 'This function requires one argument');
|
||||
assert(~isempty(var_model_name) && ischar(var_model_name), ...
|
||||
'The sole argument must be a non-empty string');
|
||||
if ~isfield(M_.var, var_model_name)
|
||||
error(['Could not find ' var_model_name ' in M_.var. ' ...
|
||||
'First declare it via the var_model statement.']);
|
||||
end
|
||||
|
||||
%% Call Dynamic Function
|
||||
[junk, g1] = feval([M_.fname '_dynamic'], ...
|
||||
ones(max(max(M_.lead_lag_incidence)), 1), ...
|
||||
ones(1, M_.exo_nbr), ...
|
||||
M_.params, ...
|
||||
zeros(M_.endo_nbr, 1), ...
|
||||
1);
|
||||
|
||||
% Choose rows of Jacobian based on equation tags
|
||||
ntags = length(M_.var.(var_model_name).eqtags);
|
||||
g1rows = zeros(ntags, 1);
|
||||
for i = 1:ntags
|
||||
idxs = strcmp(M_.equations_tags(:, 3), M_.var.(var_model_name).eqtags{i});
|
||||
if any(idxs)
|
||||
g1rows(i) = M_.equations_tags{idxs, 1};
|
||||
end
|
||||
end
|
||||
g1 = -1 * g1(g1rows, :);
|
||||
|
||||
% Check for leads
|
||||
if rows(M_.lead_lag_incidence) == 3
|
||||
idxs = M_.lead_lag_incidence(3, M_.lead_lag_incidence(3, :) ~= 0);
|
||||
assert(~any(any(g1(g1rows, idxs))), ...
|
||||
['You cannot have leads in the equations specified by ' strjoin(M_.var.(var_model_name).eqtags, ',')]);
|
||||
end
|
||||
|
||||
%% Organize AR & EC matrices
|
||||
assert(length(M_.var.(var_model_name).lhs) == rows(g1));
|
||||
|
||||
% Find RHS vars for AR & EC matrices
|
||||
arRhsVars = [];
|
||||
ecRhsVars = [];
|
||||
lhs = M_.var.(var_model_name).lhs;
|
||||
for i = 1:length(M_.var.(var_model_name).rhs.vars_at_eq)
|
||||
vars = M_.var.(var_model_name).rhs.vars_at_eq{i}.var;
|
||||
rhsvars{i}.vars = [];
|
||||
rhsvars{i}.lags = [];
|
||||
rhsvars{i}.arRhsIdxs = [];
|
||||
rhsvars{i}.ecRhsIdxs = [];
|
||||
for j = 1:length(vars)
|
||||
if vars(j) <= M_.orig_endo_nbr
|
||||
% vars(j) is not an aux var
|
||||
if ismember(vars(j), lhs)
|
||||
arRhsVars = union(arRhsVars, vars(j), 'stable');
|
||||
rhsvars{i}.arRhsIdxs = [rhsvars{i}.arRhsIdxs find(arRhsVars == vars(j))];
|
||||
rhsvars{i}.ecRhsIdxs = [rhsvars{i}.ecRhsIdxs -1];
|
||||
else
|
||||
ecRhsVars = union(ecRhsVars, vars(j), 'stable');
|
||||
rhsvars{i}.arRhsIdxs = [rhsvars{i}.arRhsIdxs -1];
|
||||
rhsvars{i}.ecRhsIdxs = [rhsvars{i}.ecRhsIdxs find(ecRhsVars == vars(j))];
|
||||
end
|
||||
else
|
||||
% Search aux vars for matching lhs var
|
||||
lhsvaridx = findLhsInAuxVar(vars(j), lhs);
|
||||
if lhsvaridx >= 1
|
||||
arRhsVars = union(arRhsVars, lhsvaridx, 'stable');
|
||||
rhsvars{i}.arRhsIdxs = [rhsvars{i}.arRhsIdxs find(arRhsVars == lhsvaridx)];
|
||||
rhsvars{i}.ecRhsIdxs = [rhsvars{i}.ecRhsIdxs -1];
|
||||
else
|
||||
% otherwise find endog that corresponds to this aux var
|
||||
varidx = findVarNoLag(vars(j));
|
||||
ecRhsVars = union(ecRhsVars, varidx, 'stable');
|
||||
rhsvars{i}.arRhsIdxs = [rhsvars{i}.arRhsIdxs -1];
|
||||
rhsvars{i}.ecRhsIdxs = [rhsvars{i}.ecRhsIdxs find(ecRhsVars == varidx)];
|
||||
end
|
||||
end
|
||||
end
|
||||
rhsvars{i}.vars = vars;
|
||||
rhsvars{i}.lags = M_.var.(var_model_name).rhs.vars_at_eq{i}.lag;
|
||||
end
|
||||
|
||||
% Initialize matrices
|
||||
oo_.var.(var_model_name).ar = zeros(length(lhs), length(arRhsVars), M_.var.(var_model_name).max_lag);
|
||||
oo_.var.(var_model_name).ec = zeros(length(lhs), length(ecRhsVars), M_.var.(var_model_name).max_lag);
|
||||
oo_.var.(var_model_name).ar_idx = arRhsVars;
|
||||
oo_.var.(var_model_name).ec_idx = ecRhsVars;
|
||||
% Fill matrices
|
||||
for i = 1:length(rhsvars)
|
||||
for j = 1:length(rhsvars{i}.vars)
|
||||
var = rhsvars{i}.vars(j);
|
||||
if rhsvars{i}.lags(j) == -1
|
||||
g1col = M_.lead_lag_incidence(1, var);
|
||||
else
|
||||
g1col = M_.lead_lag_incidence(2, var);
|
||||
end
|
||||
if g1col ~= 0 && any(g1(:, g1col))
|
||||
if rhsvars{i}.arRhsIdxs(j) > 0
|
||||
% Fill AR
|
||||
[lag, ndiffs] = findLagForVar(var, -rhsvars{i}.lags(j), 0, arRhsVars);
|
||||
oo_.var.(var_model_name).ar(i, rhsvars{i}.arRhsIdxs(j), lag) = ...
|
||||
oo_.var.(var_model_name).ar(i, rhsvars{i}.arRhsIdxs(j), lag) + g1(i, g1col);
|
||||
ndiffs = ndiffs - 1;
|
||||
if ndiffs == 1
|
||||
oo_.var.(var_model_name).ar(i, rhsvars{i}.arRhsIdxs(j), lag + 1) = ...
|
||||
oo_.var.(var_model_name).ar(i, rhsvars{i}.arRhsIdxs(j), lag + 1) - g1(i, g1col);
|
||||
elseif ndiffs > 1
|
||||
error('No support yet for more than one nested diff');
|
||||
end
|
||||
elseif rhsvars{i}.ecRhsIdxs(j) > 0
|
||||
% Fill EC
|
||||
[lag, ndiffs] = findLagForVar(var, -rhsvars{i}.lags(j), 0, ecRhsVars);
|
||||
oo_.var.(var_model_name).ec(i, rhsvars{i}.ecRhsIdxs(j), lag) = ...
|
||||
oo_.var.(var_model_name).ec(i, rhsvars{i}.ecRhsIdxs(j), lag) + g1(i, g1col);
|
||||
if ndiffs == 1
|
||||
oo_.var.(var_model_name).ec(i, rhsvars{i}.ecRhsIdxs(j), lag + 1) = ...
|
||||
oo_.var.(var_model_name).ec(i, rhsvars{i}.ecRhsIdxs(j), lag + 1) - g1(i, g1col);
|
||||
elseif ndiffs > 1
|
||||
error('No support yet for more than one nested diff');
|
||||
end
|
||||
else
|
||||
error('Shouldn''t arrive here');
|
||||
end
|
||||
end
|
||||
end
|
||||
end
|
||||
end
|
||||
|
||||
|
||||
|
||||
function lhsvaridx = findLhsInAuxVar(auxVar, lhsvars)
|
||||
|
||||
global M_
|
||||
|
||||
if auxVar <= M_.orig_endo_nbr
|
||||
lhsvaridx = -1;
|
||||
return
|
||||
end
|
||||
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == auxVar);
|
||||
|
||||
if ismember(av.orig_index, lhsvars)
|
||||
lhsvaridx = av.orig_index;
|
||||
else
|
||||
lhsvaridx = findLhsInAuxVar(av.orig_index, lhsvars);
|
||||
end
|
||||
end
|
||||
|
||||
|
||||
|
||||
function idx = findVarNoLag(auxVar)
|
||||
|
||||
global M_
|
||||
|
||||
if auxVar <= M_.orig_endo_nbr
|
||||
error('Shouldn''t arrive here')
|
||||
end
|
||||
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == auxVar);
|
||||
|
||||
if ~isempty(av.unary_op_handle)
|
||||
idx = av.endo_index;
|
||||
else
|
||||
if av.orig_index <= M_.orig_endo_nbr
|
||||
idx = av.orig_index;
|
||||
else
|
||||
idx = findVarNoLag(av.orig_index);
|
||||
end
|
||||
end
|
||||
end
|
||||
|
||||
|
||||
|
||||
function [lag, ndiffs] = findLagForVar(auxVar, lag, ndiffs, rhsVars)
|
||||
|
||||
global M_
|
||||
|
||||
if auxVar <= M_.orig_endo_nbr
|
||||
return
|
||||
end
|
||||
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == auxVar);
|
||||
|
||||
if av.type == 8
|
||||
ndiffs = ndiffs + 1;
|
||||
end
|
||||
|
||||
if ismember(av.endo_index, rhsVars)
|
||||
if ~isempty(av.unary_op_handle) && (av.type == 8 || av.type == 9)
|
||||
lag = lag + abs(av.orig_lead_lag);
|
||||
end
|
||||
elseif ismember(av.orig_index, rhsVars)
|
||||
if av.orig_index <= M_.orig_endo_nbr
|
||||
lag = lag + abs(av.orig_lead_lag);
|
||||
else
|
||||
[lag, ndiffs] = findLagForVar(av.orig_index, lag + 1, ndiffs, rhsVars);
|
||||
end
|
||||
else
|
||||
if av.type == 8
|
||||
[lag, ndiffs] = findLagForVar(av.orig_index, lag, ndiffs, rhsVars);
|
||||
else
|
||||
[lag, ndiffs] = findLagForVar(av.orig_index, lag + 1, ndiffs, rhsVars);
|
||||
end
|
||||
end
|
||||
assert(lag > 0)
|
||||
end
|
||||
|
||||
|
|
@ -1,211 +0,0 @@
|
|||
function get_ar_matrices(var_model_name)
|
||||
|
||||
% Gets the autoregressive matrices associated with the var specified by var_model_name.
|
||||
% Output stored in cellarray oo_.var.(var_model_name).AutoregressiveMatrices, with
|
||||
% oo_.var.(var_model_name).AutoregressiveMatrices{i} being the AR matrix at time t-i. Each
|
||||
% AR matrix is stored with rows organized by the ordering of the equation tags
|
||||
% found in M_.var.(var_model_name).eqtags and columns organized consistently.
|
||||
%
|
||||
% INPUTS
|
||||
%
|
||||
% var_model_name [string] the name of the VAR model
|
||||
%
|
||||
% OUTPUTS
|
||||
%
|
||||
% NONE
|
||||
|
||||
% Copyright (C) 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 <http://www.gnu.org/licenses/>.
|
||||
|
||||
global M_ oo_
|
||||
|
||||
%% Check inputs and initialize output
|
||||
assert(nargin == 1, 'This function requires one argument');
|
||||
assert(~isempty(var_model_name) && ischar(var_model_name), ...
|
||||
'The sole argument must be a non-empty string');
|
||||
if ~isfield(M_.var, var_model_name)
|
||||
error(['Could not find ' var_model_name ' in M_.var. ' ...
|
||||
'First declare it via the var_model statement.']);
|
||||
end
|
||||
|
||||
%% Call Dynamic Function
|
||||
[junk, g1] = feval([M_.fname '_dynamic'], ...
|
||||
ones(max(max(M_.lead_lag_incidence)), 1), ...
|
||||
ones(1, M_.exo_nbr), ...
|
||||
M_.params, ...
|
||||
zeros(M_.endo_nbr, 1), ...
|
||||
1);
|
||||
|
||||
% Choose rows of Jacobian based on equation tags
|
||||
ntags = length(M_.var.(var_model_name).eqtags);
|
||||
g1rows = zeros(ntags, 1);
|
||||
for i = 1:ntags
|
||||
idxs = strcmp(M_.equations_tags(:, 3), M_.var.(var_model_name).eqtags{i});
|
||||
if any(idxs)
|
||||
g1rows(i) = M_.equations_tags{idxs, 1};
|
||||
end
|
||||
end
|
||||
g1 = -1 * g1(g1rows, :);
|
||||
|
||||
% Check for leads
|
||||
if rows(M_.lead_lag_incidence) == 3
|
||||
idxs = M_.lead_lag_incidence(3, M_.lead_lag_incidence(3, :) ~= 0);
|
||||
assert(~any(any(g1(g1rows, idxs))), ...
|
||||
['You cannot have leads in the equations specified by ' strjoin(M_.var.(var_model_name).eqtags, ',')]);
|
||||
end
|
||||
|
||||
%% Organize AR matrices
|
||||
assert(length(M_.var.(var_model_name).lhs) == rows(g1));
|
||||
|
||||
% Initialize AR matrices
|
||||
% ECM
|
||||
rhsvars = [];
|
||||
rhslag = [];
|
||||
|
||||
maxlag = 0;
|
||||
for i=1:length(M_.var.(var_model_name).rhs.vars_at_eq)
|
||||
maxlag = max([maxlag; -min(M_.var.(var_model_name).rhs.vars_at_eq{i}.lag)]);
|
||||
end
|
||||
|
||||
for i = 1:length(M_.var.(var_model_name).rhs.vars_at_eq)
|
||||
rhsvars = union(rhsvars, M_.var.(var_model_name).rhs.vars_at_eq{i}.var);
|
||||
rhslag = union(rhslag, M_.var.(var_model_name).rhs.vars_at_eq{i}.lag);
|
||||
end
|
||||
if ~isempty(rhslag)
|
||||
maxlag = max(maxlag, max(abs(rhslag)));
|
||||
end
|
||||
vars = union(rhsvars, M_.var.(var_model_name).lhs);
|
||||
vars = vars(vars > M_.orig_endo_nbr);
|
||||
for i = 1:length(vars)
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == vars(i));
|
||||
if isfield(av, 'orig_lead_lag')
|
||||
maxlag = max(maxlag, abs(av.orig_lead_lag));
|
||||
end
|
||||
end
|
||||
|
||||
Bvars = setdiff(rhsvars, M_.var.(var_model_name).lhs);
|
||||
orig_diff_var_vec = M_.var.(var_model_name).orig_diff_var(M_.var.(var_model_name).diff);
|
||||
diff_vars = M_.var.(var_model_name).lhs(M_.var.(var_model_name).lhs > M_.orig_endo_nbr);
|
||||
drop_diff_avs_related_to = [];
|
||||
for i = 1:length(diff_vars)
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == diff_vars(i));
|
||||
assert(any(orig_diff_var_vec == av.orig_index));
|
||||
if av.type == 8
|
||||
drop_diff_avs_related_to = [drop_diff_avs_related_to av.orig_index];
|
||||
end
|
||||
end
|
||||
keep = true(length(Bvars), 1);
|
||||
Bvars_diff_index = zeros(length(Bvars), 1);
|
||||
for i = sum(Bvars <= M_.orig_endo_nbr)+1:length(Bvars)
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == Bvars(i));
|
||||
if av.type == 8
|
||||
if any(drop_diff_avs_related_to == av.orig_index)
|
||||
assert(any(orig_diff_var_vec == av.orig_index));
|
||||
keep(i) = false;
|
||||
else
|
||||
if any(Bvars_diff_index == av.orig_index)
|
||||
keep(i) = false;
|
||||
else
|
||||
Bvars_diff_index(i) = av.orig_index;
|
||||
end
|
||||
end
|
||||
end
|
||||
end
|
||||
Bvars = Bvars(keep);
|
||||
Bvars_diff_index = Bvars_diff_index(keep);
|
||||
oo_.var.(var_model_name).ecm_idx = Bvars;
|
||||
|
||||
% AR
|
||||
narvars = length(M_.var.(var_model_name).lhs);
|
||||
nothvars = length(Bvars);
|
||||
for i = 1:maxlag + 1
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices{i} = zeros(narvars, narvars);
|
||||
oo_.var.(var_model_name).ecm{i} = zeros(narvars, nothvars);
|
||||
end
|
||||
|
||||
ecm_assigned = false;
|
||||
for i = 1:2
|
||||
if i == 1
|
||||
baselag = 2;
|
||||
else
|
||||
baselag = 1;
|
||||
end
|
||||
for j = 1:size(M_.lead_lag_incidence, 2)
|
||||
if M_.lead_lag_incidence(i, j) ~= 0 && any(g1(:, M_.lead_lag_incidence(i, j)))
|
||||
if j > M_.orig_endo_nbr
|
||||
av = M_.aux_vars([M_.aux_vars.endo_index] == j);
|
||||
assert(~isempty(av));
|
||||
if av.type == 8
|
||||
col = M_.var.(var_model_name).orig_diff_var == av.orig_index;
|
||||
if any(col)
|
||||
assert(any(orig_diff_var_vec == av.orig_index));
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices{(av.orig_lead_lag * - 1) + baselag}(:, col) = ...
|
||||
g1(:, M_.lead_lag_incidence(i, j));
|
||||
else
|
||||
col = Bvars_diff_index == av.orig_index;
|
||||
ecm_assigned = true;
|
||||
oo_.var.(var_model_name).ecm{(av.orig_lead_lag * - 1) + baselag}(:, col) = ...
|
||||
g1(:, M_.lead_lag_incidence(i, j));
|
||||
end
|
||||
else
|
||||
col = M_.var.(var_model_name).lhs == av.orig_index;
|
||||
if any(col)
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices{(av.orig_lead_lag * - 1) + baselag}(:, col) = ...
|
||||
g1(:, M_.lead_lag_incidence(i, j));
|
||||
else
|
||||
col = Bvars == av.orig_index;
|
||||
ecm_assigned = true;
|
||||
oo_.var.(var_model_name).ecm{(av.orig_lead_lag * - 1) + baselag}(:, col) = ...
|
||||
g1(:, M_.lead_lag_incidence(i, j));
|
||||
end
|
||||
end
|
||||
else
|
||||
col = M_.var.(var_model_name).lhs == j;
|
||||
if ~any(col)
|
||||
col = Bvars == j;
|
||||
ecm_assigned = true;
|
||||
oo_.var.(var_model_name).ecm{baselag}(:, col) = ...
|
||||
g1(:, M_.lead_lag_incidence(i, j));
|
||||
else
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices{baselag}(:, col) = ...
|
||||
g1(:, M_.lead_lag_incidence(i, j));
|
||||
end
|
||||
end
|
||||
end
|
||||
end
|
||||
end
|
||||
|
||||
for i = 1:length(M_.var.(var_model_name).lhs)
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices{1}(i, M_.var.(var_model_name).lhs == M_.var.(var_model_name).lhs(i)) = ...
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices{1}(i, M_.var.(var_model_name).lhs == M_.var.(var_model_name).lhs(i)) + 1;
|
||||
end
|
||||
|
||||
if any(oo_.var.(var_model_name).AutoregressiveMatrices{1}(:))
|
||||
error('This is not a VAR model! Contemporaneous endogenous variables are not allowed.')
|
||||
end
|
||||
|
||||
% Remove time t matrix for autoregressive part
|
||||
oo_.var.(var_model_name).AutoregressiveMatrices = oo_.var.(var_model_name).AutoregressiveMatrices(2:end);
|
||||
|
||||
if ecm_assigned
|
||||
oo_.var.(var_model_name).ecm = oo_.var.(var_model_name).ecm(2:end);
|
||||
else
|
||||
% Remove error correction matrices if never assigned
|
||||
oo_.var.(var_model_name) = rmfield(oo_.var.(var_model_name), 'ecm');
|
||||
oo_.var.(var_model_name) = rmfield(oo_.var.(var_model_name), 'ecm_idx');
|
||||
end
|
||||
end
|
||||
|
|
@ -1,4 +1,5 @@
|
|||
function get_companion_matrix(var_model_name)
|
||||
%function get_companion_matrix(var_model_name)
|
||||
|
||||
% Gets the companion matrix associated with the var specified by
|
||||
% var_model_name. Output stored in cellarray oo_.var.(var_model_name).H.
|
||||
|
|
@ -28,33 +29,33 @@ function get_companion_matrix(var_model_name)
|
|||
|
||||
global oo_
|
||||
|
||||
get_ar_matrices(var_model_name);
|
||||
get_ar_ec_matrices(var_model_name);
|
||||
|
||||
% Get the number of lags
|
||||
p = length(oo_.var.(var_model_name).AutoregressiveMatrices);
|
||||
p = length(oo_.var.(var_model_name).ar);
|
||||
|
||||
% Get the number of variables
|
||||
n = length(oo_.var.(var_model_name).AutoregressiveMatrices{1});
|
||||
n = length(oo_.var.(var_model_name).ar{1});
|
||||
|
||||
if all(cellfun(@iszero, oo_.var.(var_model_name).ecm))
|
||||
% Build the companion matrix (standard VAR)
|
||||
oo_.var.(var_model_name).CompanionMatrix = zeros(n*p);
|
||||
oo_.var.(var_model_name).CompanionMatrix(1:n,1:n) = oo_.var.(var_model_name).AutoregressiveMatrices{1};
|
||||
oo_.var.(var_model_name).CompanionMatrix(1:n,1:n) = oo_.var.(var_model_name).ar{1};
|
||||
if p>1
|
||||
for i=2:p
|
||||
oo_.var.(var_model_name).CompanionMatrix(1:n,(i-1)*n+(1:n)) = oo_.var.(var_model_name).AutoregressiveMatrices{i};
|
||||
oo_.var.(var_model_name).CompanionMatrix(1:n,(i-1)*n+(1:n)) = oo_.var.(var_model_name).ar{i};
|
||||
oo_.var.(var_model_name).CompanionMatrix((i-1)*n+(1:n),(i-2)*n+(1:n)) = eye(n);
|
||||
end
|
||||
end
|
||||
else
|
||||
B = zeros(n,n,p+1);
|
||||
idx = oo_.var.(var_model_name).ecm_idx;
|
||||
B(:,:,1) = oo_.var.(var_model_name).AutoregressiveMatrices{1};
|
||||
B(:,:,1) = oo_.var.(var_model_name).ar{1};
|
||||
B(idx, idx, 1) = B(idx,idx, 1) + eye(length(idx));
|
||||
for i=2:p
|
||||
B(idx,idx,i) = oo_.var.(var_model_name).AutoregressiveMatrices{i}(idx,idx)-oo_.var.(var_model_name).AutoregressiveMatrices{i-1}(idx,idx);
|
||||
B(idx,idx,i) = oo_.var.(var_model_name).ar{i}(idx,idx)-oo_.var.(var_model_name).ar{i-1}(idx,idx);
|
||||
end
|
||||
B(idx,idx,p+1) = -oo_.var.(var_model_name).AutoregressiveMatrices{p}(idx,idx);
|
||||
B(idx,idx,p+1) = -oo_.var.(var_model_name).ar{p}(idx,idx);
|
||||
% Build the companion matrix (VECM, rewrite in levels)
|
||||
oo_.var.(var_model_name).CompanionMatrix = zeros(n*(p+1));
|
||||
for i=1:p
|
||||
|
|
|
|||
|
|
@ -1 +1 @@
|
|||
Subproject commit 912261e5fcfc98db4ee4f165bbbb8f9bb4fe81b4
|
||||
Subproject commit 32041cf81a2ac76fe653bafbc71d9db536198496
|
||||
Loading…
Reference in New Issue