stepan
/
dynare
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Get rid of dsge_likelihood_hh.m in mode_compute=5

time-shift
Marco Ratto 2012-04-27 15:50:12 +02:00 committed by Michel Juillard
parent bb02c4d6df
commit 3332a7f794
3 changed files with 6 additions and 468 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

455
matlab/dsge_likelihood_hh.m
View File

@ -1,455 +0,0 @@
function [fval,llik,cost_flag,ys,trend_coeff,info] = dsge_likelihood_hh(xparam1,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults)
% function [fval,llik,cost_flag,ys,trend_coeff,info] = DsgeLikelihood_hh(xparam1,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults)
% Evaluates the posterior kernel of a dsge model.
%
% INPUTS
% xparam1 [double] vector of model parameters.
% gend [integer] scalar specifying the number of observations.
% data [double] matrix of data
% data_index [cell] cell of column vectors
% number_of_observations [integer]
% no_more_missing_observations [integer]
% OUTPUTS
% fval : value of the posterior kernel at xparam1.
% llik : probabilities at each time point
% cost_flag : zero if the function returns a penalty, one otherwise.
% ys : steady state of original endogenous variables
% trend_coeff :
% info : vector of informations about the penalty:
% 41: one (many) parameter(s) do(es) not satisfied the lower bound
% 42: one (many) parameter(s) do(es) not satisfied the upper bound
%
% SPECIAL REQUIREMENTS
%
% Copyright (C) 2004-2011 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/>.
% Declaration of the penalty as a persistent variable.
% Persistent variable 'penalty' is used to compute an endogenous penalty to
% the value 'fval' when various conditions are encountered. These conditions
% set also 'exit_flag' equal to 0 instead of 1. It is only when
% dsge_likelihood_hh() is called by an newrat() called by
% dynare_estimation_1() that 'exit_flag' is ignored and penalized 'fval' is
% actually used.
% In that case, 'penalty' is properly initialized, at the very end of the
% present function, by a call to dsge_likelihood_hh() made in
% initial_estimation_checks(). If a condition triggers exit_flag ==
% 0, initial_estimation_checks() triggers an error.
% In summary, an initial call to the present function, without triggering
% any condition, guarantees that 'penalty' is properly initialized when needed.
persistent penalty
% Initialization of the returned variables and others...
fval = [];
ys = [];
trend_coeff = [];
cost_flag = 1;
llik=NaN;
info = 0;
singularity_flag = 0;
if DynareOptions.block == 1
error('DsgeLikelihood_hh:: This routine (called if mode_compute==5) is not compatible with the block option!')
end
%------------------------------------------------------------------------------
% 1. Get the structural parameters & define penalties
%------------------------------------------------------------------------------
% Return, with endogenous penalty, if some parameters are smaller than the lower bound of the prior domain.
if ~isequal(DynareOptions.mode_compute,1) && any(xparam1<BayesInfo.lb)
k = find(xparam1<BayesInfo.lb);
fval = penalty+sum((BayesInfo.lb(k)-xparam1(k)).^2);
exit_flag = 0;
info = 41;
return
end
% Return, with endogenous penalty, if some parameters are greater than the upper bound of the prior domain.
if ~isequal(DynareOptions.mode_compute,1) && any(xparam1>BayesInfo.ub)
k = find(xparam1>BayesInfo.ub);
fval = penalty+sum((xparam1(k)-BayesInfo.ub(k)).^2);
exit_flag = 0;
info = 42;
return
end
% Get the diagonal elements of the covariance matrices for the structural innovations (Q) and the measurement error (H).
Q = Model.Sigma_e;
H = Model.H;
for i=1:EstimatedParameters.nvx
k =EstimatedParameters.var_exo(i,1);
Q(k,k) = xparam1(i)*xparam1(i);
end
offset = EstimatedParameters.nvx;
if EstimatedParameters.nvn
for i=1:EstimatedParameters.nvn
k = EstimatedParameters.var_endo(i,1);
H(k,k) = xparam1(i+offset)*xparam1(i+offset);
end
offset = offset+EstimatedParameters.nvn;
else
H = zeros(DynareDataset.info.nvobs);
end
% Get the off-diagonal elements of the covariance matrix for the structural innovations. Test if Q is positive definite.
if EstimatedParameters.ncx
for i=1:EstimatedParameters.ncx
k1 =EstimatedParameters.corrx(i,1);
k2 =EstimatedParameters.corrx(i,2);
Q(k1,k2) = xparam1(i+offset)*sqrt(Q(k1,k1)*Q(k2,k2));
Q(k2,k1) = Q(k1,k2);
end
% Try to compute the cholesky decomposition of Q (possible iff Q is positive definite)
[CholQ,testQ] = chol(Q);
if testQ
% The variance-covariance matrix of the structural innovations is not definite positive. We have to compute the eigenvalues of this matrix in order to build the endogenous penalty.
a = diag(eig(Q));
k = find(a < 0);
if k > 0
fval = penalty+sum(-a(k));
exit_flag = 0;
info = 43;
return
end
end
offset = offset+EstimatedParameters.ncx;
end
% Get the off-diagonal elements of the covariance matrix for the measurement errors. Test if H is positive definite.
if EstimatedParameters.ncn
for i=1:EstimatedParameters.ncn
k1 = DynareOptions.lgyidx2varobs(EstimatedParameters.corrn(i,1));
k2 = DynareOptions.lgyidx2varobs(EstimatedParameters.corrn(i,2));
H(k1,k2) = xparam1(i+offset)*sqrt(H(k1,k1)*H(k2,k2));
H(k2,k1) = H(k1,k2);
end
% Try to compute the cholesky decomposition of H (possible iff H is positive definite)
[CholH,testH] = chol(H);
if testH
% The variance-covariance matrix of the measurement errors is not definite positive. We have to compute the eigenvalues of this matrix in order to build the endogenous penalty.
a = diag(eig(H));
k = find(a < 0);
if k > 0
fval = penalty+sum(-a(k));
exit_flag = 0;
info = 44;
return
end
end
offset = offset+EstimatedParameters.ncn;
end
% Update estimated structural parameters in Mode.params.
if EstimatedParameters.np > 0
Model.params(EstimatedParameters.param_vals(:,1)) = xparam1(offset+1:end);
end
% Update Model.Sigma_e and Model.H.
Model.Sigma_e = Q;
Model.H = H;
%------------------------------------------------------------------------------
% 2. call model setup & reduction program
%------------------------------------------------------------------------------
% Linearize the model around the deterministic sdteadystate and extract the matrices of the state equation (T and R).
[T,R,SteadyState,info,Model,DynareOptions,DynareResults] = dynare_resolve(Model,DynareOptions,DynareResults,'restrict');
% Return, with endogenous penalty when possible, if dynare_resolve issues an error code (defined in resol).
if info(1) == 1 || info(1) == 2 || info(1) == 5 || info(1) == 22 || info(1) == 24
fval = penalty+1;
info = info(1);
exit_flag = 0;
return
elseif info(1) == 3 || info(1) == 4 || info(1)==6 ||info(1) == 19 || info(1) == 20 || info(1) == 21 || info(1) == 23
fval = penalty+info(2);
info = info(1);
exit_flag = 0;
return
end
% Define a vector of indices for the observed variables. Is this really usefull?...
BayesInfo.mf = BayesInfo.mf1;
% Define the constant vector of the measurement equation.
if DynareOptions.noconstant
constant = zeros(DynareDataset.info.nvobs,1);
else
if DynareOptions.loglinear
constant = log(SteadyState(BayesInfo.mfys));
else
constant = SteadyState(BayesInfo.mfys);
end
end
% Define the deterministic linear trend of the measurement equation.
if BayesInfo.with_trend
trend_coeff = zeros(DynareDataset.info.nvobs,1);
t = DynareOptions.trend_coeffs;
for i=1:length(t)
if ~isempty(t{i})
trend_coeff(i) = evalin('base',t{i});
end
end
trend = repmat(constant,1,DynareDataset.info.ntobs)+trend_coeff*[1:DynareDataset.info.ntobs];
else
trend = repmat(constant,1,DynareDataset.info.ntobs);
end
% Get needed informations for kalman filter routines.
start = DynareOptions.presample+1;
Z = BayesInfo.mf; % old mf
no_missing_data_flag = ~DynareDataset.missing.state;
mm = length(T); % old np
pp = DynareDataset.info.nvobs;
rr = length(Q);
kalman_tol = DynareOptions.kalman_tol;
riccati_tol = DynareOptions.riccati_tol;
Y = DynareDataset.data-trend;
%------------------------------------------------------------------------------
% 3. Initial condition of the Kalman filter
%------------------------------------------------------------------------------
kalman_algo = DynareOptions.kalman_algo;
% resetting measurement errors covariance matrix for univariate filters
if (kalman_algo == 2) || (kalman_algo == 4)
if isequal(H,0)
H = zeros(nobs,1);
mmm = mm;
else
if all(all(abs(H-diag(diag(H)))<1e-14))% ie, the covariance matrix is diagonal...
H = diag(H);
mmm = mm;
else
Z = [Z, eye(pp)];
T = blkdiag(T,zeros(pp));
Q = blkdiag(Q,H);
R = blkdiag(R,eye(pp));
Pstar = blkdiag(Pstar,H);
Pinf = blckdiag(Pinf,zeros(pp));
H = zeros(nobs,1);
mmm = mm+pp;
end
end
end
diffuse_periods = 0;
correlated_errors_have_been_checked = 0;
singular_diffuse_filter = 0;
switch DynareOptions.lik_init
case 1% Standard initialization with the steady state of the state equation.
if kalman_algo~=2
% Use standard kalman filter except if the univariate filter is explicitely choosen.
kalman_algo = 1;
end
Pstar = lyapunov_symm(T,R*Q*R',DynareOptions.qz_criterium,DynareOptions.lyapunov_complex_threshold);
Pinf = [];
a = zeros(mm,1);
Zflag = 0;
case 2% Initialization with large numbers on the diagonal of the covariance matrix if the states (for non stationary models).
if kalman_algo ~= 2
% Use standard kalman filter except if the univariate filter is explicitely choosen.
kalman_algo = 1;
end
Pstar = DynareOptions.Harvey_scale_factor*eye(mm);
Pinf = [];
a = zeros(mm,1);
Zflag = 0;
case 3% Diffuse Kalman filter (Durbin and Koopman)
% Use standard kalman filter except if the univariate filter is explicitely choosen.
if kalman_algo == 0
kalman_algo = 3;
elseif ~((kalman_algo == 3) || (kalman_algo == 4))
error(['diffuse filter: options_.kalman_algo can only be equal ' ...
'to 0 (default), 3 or 4'])
end
[Z,T,R,QT,Pstar,Pinf] = schur_statespace_transformation(Z,T,R,Q,DynareOptions.qz_criterium);
Zflag = 1;
% Run diffuse kalman filter on first periods.
if (kalman_algo==3)
% Multivariate Diffuse Kalman Filter
if no_missing_data_flag
[dLIK,dlik,a,Pstar] = kalman_filter_d(Y, 1, size(Y,2), ...
zeros(mm,1), Pinf, Pstar, ...
kalman_tol, riccati_tol, DynareOptions.presample, ...
T,R,Q,H,Z,mm,pp,rr);
else
[dLIK,dlik,a,Pstar] = missing_observations_kalman_filter_d(DynareDataset.missing.aindex,DynareDataset.missing.number_of_observations,DynareDataset.missing.no_more_missing_observations, ...
Y, 1, size(Y,2), ...
zeros(mm,1), Pinf, Pstar, ...
kalman_tol, riccati_tol, DynareOptions.presample, ...
T,R,Q,H,Z,mm,pp,rr);
end
diffuse_periods = length(dlik);
if isinf(dLIK)
% Go to univariate diffuse filter if singularity problem.
singular_diffuse_filter
end
end
if singular_diffuse_filter || (kalman_algo==4)
% Univariate Diffuse Kalman Filter
if isequal(H,0)
H1 = zeros(nobs,1);
mmm = mm;
else
if all(all(abs(H-diag(diag(H)))<1e-14))% ie, the covariance matrix is diagonal...
H1 = diag(H);
mmm = mm;
else
Z = [Z, eye(pp)];
T = blkdiag(T,zeros(pp));
Q = blkdiag(Q,H);
R = blkdiag(R,eye(pp));
Pstar = blkdiag(Pstar,H);
Pinf = blckdiag(Pinf,zeros(pp));
H1 = zeros(nobs,1);
mmm = mm+pp;
end
end
% no need to test again for correlation elements
correlated_errors_have_been_checked = 1;
[dLIK,dlik,a,Pstar] = univariate_kalman_filter_d(DynareDataset.missing.aindex,...
DynareDataset.missing.number_of_observations,...
DynareDataset.missing.no_more_missing_observations, ...
Y, 1, size(Y,2), ...
zeros(mmm,1), Pinf, Pstar, ...
kalman_tol, riccati_tol, DynareOptions.presample, ...
T,R,Q,H1,Z,mmm,pp,rr);
diffuse_periods = length(dlik);
end
case 4% Start from the solution of the Riccati equation.
if kalman_algo ~= 2
kalman_algo = 1;
end
if isequal(H,0)
[err,Pstar] = kalman_steady_state(transpose(T),R*Q*transpose(R),transpose(build_selection_matrix(Z,np,length(Z))));
else
[err,Pstar] = kalman_steady_state(transpose(T),R*Q*transpose(R),transpose(build_selection_matrix(Z,np,length(Z))),H);
end
if err
disp(['DsgeLikelihood:: I am not able to solve the Riccati equation, so I switch to lik_init=1!']);
DynareOptions.lik_init = 1;
Pstar = lyapunov_symm(T,R*Q*R',DynareOptions.qz_criterium,DynareOptions.lyapunov_complex_threshold);
end
Pinf = [];
otherwise
error('DsgeLikelihood:: Unknown initialization approach for the Kalman filter!')
end
%------------------------------------------------------------------------------
% 4. Likelihood evaluation
%------------------------------------------------------------------------------
if ((kalman_algo==1) || (kalman_algo==3))% Multivariate Kalman Filter
if no_missing_data_flag
[LIK,lik] = kalman_filter(Y,diffuse_periods+1,size(Y,2), ...
a,Pstar, ...
kalman_tol, riccati_tol, ...
DynareOptions.presample, ...
T,Q,R,H,Z,mm,pp,rr,Zflag,diffuse_periods);
else
[LIK,lik] = missing_observations_kalman_filter(DynareDataset.missing.aindex,DynareDataset.missing.number_of_observations,DynareDataset.missing.no_more_missing_observations,Y,diffuse_periods+1,size(Y,2), ...
a, Pstar, ...
kalman_tol, DynareOptions.riccati_tol, ...
DynareOptions.presample, ...
T,Q,R,H,Z,mm,pp,rr,Zflag,diffuse_periods);
end
if isinf(LIK)
if kalman_algo == 1
kalman_algo = 2;
else
kalman_algo = 4;
end
singularity_flag = 1;
else
if DynareOptions.lik_init==3
LIK = LIK + dLIK;
lik = [dlik; lik];
end
end
end
if (kalman_algo==2) || (kalman_algo==4)
% Univariate Kalman Filter
% resetting measurement error covariance matrix when necessary %
if ~correlated_errors_have_been_checked
if isequal(H,0)
H = zeros(nobs,1);
mmm = mm;
else
if all(all(abs(H-diag(diag(H)))<1e-14))% ie, the covariance matrix is diagonal...
H = diag(H);
mmm = mm;
else
Z = [Z, eye(pp)];
T = blkdiag(T,zeros(pp));
Q = blkdiag(Q,H);
R = blkdiag(R,eye(pp));
Pstar = blkdiag(Pstar,H);
Pinf = blckdiag(Pinf,zeros(pp));
H = zeros(nobs,1);
mmm = mm+pp;
end
end
end
[LIK,lik] = univariate_kalman_filter(DynareDataset.missing.aindex,DynareDataset.missing.number_of_observations,DynareDataset.missing.no_more_missing_observations,Y,diffuse_periods+1,size(Y,2), ...
a,Pstar, ...
DynareOptions.kalman_tol, ...
DynareOptions.riccati_tol, ...
DynareOptions.presample, ...
T,Q,R,H,Z,mmm,pp,rr,diffuse_periods);
if DynareOptions.lik_init==3
LIK = LIK+dLIK;
lik = [dlik; lik];
end
end
if isnan(LIK)
info = 45;
exit_flag = 0;
return
end
if imag(LIK)~=0
likelihood = penalty;
else
likelihood = LIK;
end
% ------------------------------------------------------------------------------
% 5. Adds prior if necessary
% ------------------------------------------------------------------------------
lnprior = priordens(xparam1,BayesInfo.pshape,BayesInfo.p6,BayesInfo.p7,BayesInfo.p3,BayesInfo.p4);
fval = (likelihood-lnprior);
% Update DynareOptions.kalman_algo.
DynareOptions.kalman_algo = kalman_algo;
% Update the penalty.
penalty = fval;
% Add the prior density at the top of the vector for the density of each observation.
lik=lik(start:end,:);
llik=[-lnprior; lik(:)];

7
matlab/initial_estimation_checks.m
View File

@ -47,12 +47,7 @@ else
end
if DynareOptions.mode_compute==5
if strcmp(func2str(objective_function),'dsge_likelihood')
% this call is necessary to initialized persistent variable
% 'penalty' in dsge_likelihood_hh
[fval,llik,cost_flag,ys,trend_coeff,info] = ...
dsge_likelihood_hh(xparam1,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
else
if ~strcmp(func2str(objective_function),'dsge_likelihood')
error('Options mode_compute=5 is not compatible with non linear filters or Dsge-VAR models!')
end
end

12
matlab/newrat.m
View File

@ -54,7 +54,8 @@ htol=htol_base;
htol0=htol_base;
gibbstol=length(BayesInfo.pshape)/50; %25;
func_hh = str2func([func2str(func0),'_hh']);
% func0 = str2func([func2str(func0),'_hh']);
% func0 = func0;
fval0=feval(func0,x,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
fval=fval0;
@ -62,7 +63,7 @@ fval=fval0;
mr_hessian(1,x,[],[],[],DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
if isempty(hh)
[dum, gg, htol0, igg, hhg, h1]=mr_hessian(0,x,func_hh,flagit,htol,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
[dum, gg, htol0, igg, hhg, h1]=mr_hessian(0,x,func0,flagit,htol,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
hh0 = reshape(dum,nx,nx);
hh=hhg;
if min(eig(hh0))<0
@ -155,7 +156,7 @@ while norm(gg)>gtol && check==0 && jit<nit
if flagit==2
hh=hh0;
elseif flagg>0
[dum, gg, htol0, igg, hhg,h1]=mr_hessian(0,xparam1,func_hh,flagg,ftol0,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
[dum, gg, htol0, igg, hhg,h1]=mr_hessian(0,xparam1,func0,flagg,ftol0,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
if flagg==2
hh = reshape(dum,nx,nx);
ee=eig(hh);
@ -190,7 +191,7 @@ while norm(gg)>gtol && check==0 && jit<nit
catch
save m1.mat x fval0 nig
end
[dum, gg, htol0, igg, hhg, h1]=mr_hessian(0,xparam1,func_hh,flagit,htol,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
[dum, gg, htol0, igg, hhg, h1]=mr_hessian(0,xparam1,func0,flagit,htol,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);
if htol0>htol
htol=htol0;
disp(' ')
@ -247,6 +248,3 @@ end
return
function f00 = lsearch(lam,func,x,dx,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults)
x0=x-dx*lam;
f00=feval(func,x0,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults);