dynare/matlab/identification_checks.m

132 lines
4.2 KiB
Matlab

function [condJ, ind0, indnoJ, ixnoJ, McoJ, PcoJ, jweak, jweak_pair] = identification_checks(JJ, hess_flag)
% function [condJ, ind0, indnoJ, ixnoJ, McoJ, PcoJ, jweak, jweak_pair] = identification_checks(JJ, hess_flag)
% checks for identification
%
% INPUTS
% o JJ [matrix] [output x nparams] IF hess_flag==0
% derivatives of output w.r.t. parameters and shocks
% o JJ [matrix] [nparams x nparams] IF hess_flag==1
% information matrix
%
% OUTPUTS
% o cond condition number of JJ
% o ind0 [array] binary indicator for non-zero columns of H
% o indnoJ [matrix] index of non-identified params
% o ixnoJ number of rows in indnoJ
% o Mco [array] multicollinearity coefficients
% o Pco [matrix] pairwise correlations
% o jweak [binary array] gives 1 if the parameter has Mco=1(with tolerance 1.e-10)
% o jweak_pair [binary matrix] gives 1 if a couple parameters has Pco=1(with tolerance 1.e-10)
%
% SPECIAL REQUIREMENTS
% None
% Copyright (C) 2008-2011 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/>.
% My suggestion is to have the following steps for identification check in
% dynare:
% 1. check rank of JJ at theta
npar = size(JJ,2);
indnoJ = zeros(1,npar);
ind1 = find(vnorm(JJ)>=eps); % take non-zero columns
JJ1 = JJ(:,ind1);
[eu,ee2,ee1] = svd( JJ1, 0 );
condJ= cond(JJ1);
rankJ = rank(JJ./norm(JJ),1.e-10);
rankJJ = rankJ;
% if hess_flag==0,
% rankJJ = rank(JJ'*JJ);
% end
ind0 = zeros(1,npar);
ind0(ind1) = 1;
if hess_flag==0,
% find near linear dependence problems:
McoJ = NaN(npar,1);
for ii = 1:size(JJ1,2);
McoJ(ind1(ii),:) = cosn([JJ1(:,ii),JJ1(:,find([1:1:size(JJ1,2)]~=ii))]);
end
else
deltaJ = sqrt(diag(JJ(ind1,ind1)));
tildaJ = JJ(ind1,ind1)./((deltaJ)*(deltaJ'));
McoJ(ind1,1)=(1-1./diag(inv(tildaJ)));
rhoM=sqrt(1-McoJ);
% PcoJ=inv(tildaJ);
PcoJ=NaN(npar,npar);
PcoJ(ind1,ind1)=inv(JJ(ind1,ind1));
sd=sqrt(diag(PcoJ));
PcoJ = abs(PcoJ./((sd)*(sd')));
end
ixnoJ = 0;
if rankJ<npar || rankJJ<npar || min(1-McoJ)<1.e-10
% - find out which parameters are involved
% disp('Some parameters are NOT identified by the moments included in J')
% disp(' ')
if length(ind1)<npar,
% parameters with zero column in JJ
ixnoJ = ixnoJ + 1;
indnoJ(ixnoJ,:) = (~ismember([1:npar],ind1));
end
ee0 = [rankJJ+1:length(ind1)];
for j=1:length(ee0),
% linearely dependent parameters in JJ
ixnoJ = ixnoJ + 1;
indnoJ(ixnoJ,ind1) = (abs(ee1(:,ee0(j))) > 1.e-3)';
end
else %rank(J)==length(theta) =>
% disp('All parameters are identified at theta by the moments included in J')
end
% here there is no exact linear dependence, but there are several
% near-dependencies, mostly due to strong pairwise colliniearities, which can
% be checked using
jweak=zeros(1,npar);
jweak_pair=zeros(npar,npar);
if hess_flag==0,
PcoJ = NaN(npar,npar);
for ii = 1:size(JJ1,2);
PcoJ(ind1(ii),ind1(ii)) = 1;
for jj = ii+1:size(JJ1,2);
PcoJ(ind1(ii),ind1(jj)) = cosn([JJ1(:,ii),JJ1(:,jj)]);
PcoJ(ind1(jj),ind1(ii)) = PcoJ(ind1(ii),ind1(jj));
end
end
for j=1:npar,
if McoJ(j)>(1-1.e-10),
jweak(j)=1;
[ipair, jpair] = find(PcoJ(j,j+1:end)>(1-1.e-10));
for jx=1:length(jpair),
jweak_pair(j, jpair(jx)+j)=1;
jweak_pair(jpair(jx)+j, j)=1;
end
end
end
end
jweak_pair=dyn_vech(jweak_pair)';