Merge branch 'master' of ssh://kirikou/srv/d_kirikou/git/dynare
Sébastien Villemot
commit
8236361d41
|
|
@ -53,7 +53,7 @@ elseif k <= (nvx+nvn)
|
|||
vname = deblank(options_.varobs(estim_params_.var_endo(k-estim_params_.nvx,1),:));
|
||||
nam=['SE_EOBS_',vname];
|
||||
if TeX
|
||||
tname = deblank(options_.TeX_varobs(estim_params_.var_endo(k-estim_params_.nvx,1),:));
|
||||
tname = deblank(M_.endo_names_tex(estim_params_.var_endo(k-estim_params_.nvx,1),:));
|
||||
texnam = ['$ SE_{' tname '} $'];
|
||||
end
|
||||
elseif k <= (nvx+nvn+ncx)
|
||||
|
|
|
|||
|
|
@ -1,120 +1,124 @@
|
|||
function dr=mult_elimination(varlist,M_, options_, oo_)
|
||||
% function mult_elimination()
|
||||
% replaces Lagrange multipliers in Ramsey policy by lagged value of state
|
||||
% and shock variables
|
||||
%
|
||||
% INPUT
|
||||
% none
|
||||
%
|
||||
% OUTPUT
|
||||
% dr: a structure with the new decision rule
|
||||
%
|
||||
% SPECIAL REQUIREMENTS
|
||||
% none
|
||||
|
||||
% Copyright (C) 2003-2008 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/>.
|
||||
|
||||
dr = oo_.dr;
|
||||
|
||||
nstatic = dr.nstatic;
|
||||
npred = dr.npred;
|
||||
order_var = dr.order_var;
|
||||
nstates = M_.endo_names(order_var(nstatic+(1:npred)),:);
|
||||
|
||||
il = strmatch('mult_',nstates);
|
||||
nil = setdiff(1:dr.npred,il);
|
||||
m_nbr = length(il);
|
||||
nm_nbr = length(nil);
|
||||
|
||||
AA1 = dr.ghx(:,nil);
|
||||
AA2 = dr.ghx(:,il);
|
||||
A1 = dr.ghx(nstatic+(1:npred),nil);
|
||||
A2 = dr.ghx(nstatic+(1:npred),il);
|
||||
B = dr.ghu(nstatic+(1:npred),:);
|
||||
A11 = A1(nil,:);
|
||||
A21 = A1(il,:);
|
||||
A12 = A2(nil,:);
|
||||
A22 = A2(il,:);
|
||||
|
||||
[Q1,R1,E1] = qr(A2);
|
||||
n1 = sum(abs(diag(R1)) > 1e-8);
|
||||
|
||||
Q1_12 = Q1(1:nm_nbr,n1+1:end);
|
||||
Q1_22 = Q1(nm_nbr+1:end,n1+1:end);
|
||||
[Q2,R2,E2] = qr(Q1_22');
|
||||
n2 = sum(abs(diag(R2)) > 1e-8);
|
||||
|
||||
R2_1 = inv(R2(1:n2,1:n2));
|
||||
|
||||
M1(order_var,:) = AA1 - AA2*E2*[R2_1*Q2(:,1:n2)'*Q1_12'; zeros(m_nbr-n2,nm_nbr)];
|
||||
M2(order_var,:) = AA2*E2*[R2_1*Q2(:,1:n2)'*[Q1_12' Q1_22']*A1; zeros(m_nbr-n2,length(nil))];
|
||||
M3(order_var,:) = dr.ghu;
|
||||
M4(order_var,:) = AA2*E2*[R2_1*Q2(:,1:n2)'*[Q1_12' Q1_22']*B; zeros(m_nbr-n2,size(B,2))];
|
||||
|
||||
endo_nbr = M_.orig_model.endo_nbr;
|
||||
exo_nbr = M_.exo_nbr;
|
||||
|
||||
lead_lag_incidence = M_.lead_lag_incidence(:,1:endo_nbr+exo_nbr);
|
||||
lead_lag_incidence1 = lead_lag_incidence(1,:) > 0;
|
||||
maximum_lag = M_.maximum_lag;
|
||||
for i=1:maximum_lag-1
|
||||
lead_lag_incidence1 = [lead_lag_incidence1; lead_lag_incidence(i,:)| ...
|
||||
lead_lag_incidence(i+1,:)];
|
||||
end
|
||||
lead_lag_incidence1 = [lead_lag_incidence1; ...
|
||||
lead_lag_incidence(M_.maximum_lag,:) > 0];
|
||||
k = find(lead_lag_incidence1');
|
||||
lead_lag_incidence1 = zeros(size(lead_lag_incidence1'));
|
||||
lead_lag_incidence1(k) = 1:length(k);
|
||||
lead_lag_incidence1 = lead_lag_incidence1';
|
||||
|
||||
kstate = zeros(0,2);
|
||||
for i=maximum_lag:-1:1
|
||||
k = find(lead_lag_incidence(i,:));
|
||||
kstate = [kstate; [k' repmat(i+1,length(k),1)]];
|
||||
end
|
||||
|
||||
dr.M1 = M1;
|
||||
dr.M2 = M2;
|
||||
dr.M3 = M3;
|
||||
dr.M4 = M4;
|
||||
|
||||
nvar = length(varlist);
|
||||
nspred = dr.nspred;
|
||||
nspred = 6;
|
||||
if nvar > 0
|
||||
res_table = zeros(2*(nspred+M_.exo_nbr),nvar);
|
||||
headers = 'Variables';
|
||||
for i=1:length(varlist)
|
||||
k = strmatch(varlist{i},M_.endo_names(dr.order_var,:),'exact');
|
||||
headers = strvcat(headers,varlist{i});
|
||||
|
||||
res_table(1:nspred,i) = M1(k,:)';
|
||||
res_table(nspred+(1:nspred),i) = M2(k,:)';
|
||||
res_table(2*nspred+(1:M_.exo_nbr),i) = M3(k,:)';
|
||||
res_table(2*nspred+M_.exo_nbr+(1:M_.exo_nbr),i) = M4(k,:)';
|
||||
end
|
||||
|
||||
my_title='ELIMINATION OF THE MULTIPLIERS';
|
||||
lab1 = M_.endo_names(dr.order_var(dr.nstatic+[ 1 2 5:8]),:);
|
||||
labels = strvcat(lab1,lab1,M_.exo_names,M_.exo_names);
|
||||
lh = size(labels,2)+2;
|
||||
dyntable(my_title,headers,labels,res_table,lh,10,6);
|
||||
disp(' ')
|
||||
end
|
||||
|
||||
function dr=mult_elimination(varlist,M_, options_, oo_)
|
||||
% function mult_elimination()
|
||||
% replaces Lagrange multipliers in Ramsey policy by lagged value of state
|
||||
% and shock variables
|
||||
%
|
||||
% INPUT
|
||||
% none
|
||||
%
|
||||
% OUTPUT
|
||||
% dr: a structure with the new decision rule
|
||||
%
|
||||
% SPECIAL REQUIREMENTS
|
||||
% none
|
||||
|
||||
% Copyright (C) 2003-2010 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/>.
|
||||
|
||||
dr = oo_.dr;
|
||||
|
||||
nstatic = dr.nstatic;
|
||||
npred = dr.npred;
|
||||
order_var = dr.order_var;
|
||||
nstates = M_.endo_names(order_var(nstatic+(1:npred)),:);
|
||||
|
||||
il = strmatch('mult_',nstates);
|
||||
nil = setdiff(1:dr.npred,il);
|
||||
m_nbr = length(il);
|
||||
nm_nbr = length(nil);
|
||||
|
||||
AA1 = dr.ghx(:,nil);
|
||||
AA2 = dr.ghx(:,il);
|
||||
A1 = dr.ghx(nstatic+(1:npred),nil);
|
||||
A2 = dr.ghx(nstatic+(1:npred),il);
|
||||
B = dr.ghu(nstatic+(1:npred),:);
|
||||
A11 = A1(nil,:);
|
||||
A21 = A1(il,:);
|
||||
A12 = A2(nil,:);
|
||||
A22 = A2(il,:);
|
||||
B1 = B(nil,:);
|
||||
B2 = B(il,:);
|
||||
|
||||
[Q1,R1,E1] = qr([A12; A22]);
|
||||
n1 = sum(abs(diag(R1)) > 1e-8);
|
||||
|
||||
Q1_12 = Q1(1:nm_nbr,n1+1:end);
|
||||
Q1_22 = Q1(nm_nbr+(1:m_nbr),n1+1:end);
|
||||
[Q2,R2,E2] = qr(Q1_22');
|
||||
n2 = sum(abs(diag(R2)) > 1e-8);
|
||||
|
||||
R2_1 = inv(R2(1:n2,1:n2));
|
||||
|
||||
M1 = AA1 - AA2*E2*[R2_1*Q2(:,1:n2)'*Q1_12'; zeros(m_nbr-n2,nm_nbr)];
|
||||
M2 = AA2*E2*[R2_1*Q2(:,1:n2)'*[Q1_12' Q1_22']*[A11;A21]; zeros(m_nbr-n2,length(nil))];
|
||||
M3 = dr.ghu;
|
||||
M4 = AA2*E2*[R2_1*Q2(:,1:n2)'*[Q1_12' Q1_22']*[B1;B2]; zeros(m_nbr-n2,size(B,2))];
|
||||
|
||||
k1 = nstatic+(1:npred);
|
||||
k1 = k1(nil);
|
||||
|
||||
endo_nbr = M_.orig_model.endo_nbr;
|
||||
exo_nbr = M_.exo_nbr;
|
||||
|
||||
lead_lag_incidence = M_.lead_lag_incidence(:,1:endo_nbr+exo_nbr);
|
||||
lead_lag_incidence1 = lead_lag_incidence(1,:) > 0;
|
||||
maximum_lag = M_.maximum_lag;
|
||||
for i=1:maximum_lag-1
|
||||
lead_lag_incidence1 = [lead_lag_incidence1; lead_lag_incidence(i,:)| ...
|
||||
lead_lag_incidence(i+1,:)];
|
||||
end
|
||||
lead_lag_incidence1 = [lead_lag_incidence1; ...
|
||||
lead_lag_incidence(M_.maximum_lag,:) > 0];
|
||||
k = find(lead_lag_incidence1');
|
||||
lead_lag_incidence1 = zeros(size(lead_lag_incidence1'));
|
||||
lead_lag_incidence1(k) = 1:length(k);
|
||||
lead_lag_incidence1 = lead_lag_incidence1';
|
||||
|
||||
kstate = zeros(0,2);
|
||||
for i=maximum_lag:-1:1
|
||||
k = find(lead_lag_incidence(i,:));
|
||||
kstate = [kstate; [k' repmat(i+1,length(k),1)]];
|
||||
end
|
||||
|
||||
dr.M1 = M1;
|
||||
dr.M2 = M2;
|
||||
dr.M3 = M3;
|
||||
dr.M4 = M4;
|
||||
|
||||
nvar = length(varlist);
|
||||
nspred = dr.nspred;
|
||||
nspred = 6;
|
||||
if nvar > 0
|
||||
res_table = zeros(2*(nspred+M_.exo_nbr),nvar);
|
||||
headers = 'Variables';
|
||||
for i=1:length(varlist)
|
||||
k = strmatch(varlist{i},M_.endo_names(dr.order_var,:),'exact');
|
||||
headers = strvcat(headers,varlist{i});
|
||||
|
||||
res_table(1:nspred,i) = M1(k,:)';
|
||||
res_table(nspred+(1:nspred),i) = M2(k,:)';
|
||||
res_table(2*nspred+(1:M_.exo_nbr),i) = M3(k,:)';
|
||||
res_table(2*nspred+M_.exo_nbr+(1:M_.exo_nbr),i) = M4(k,:)';
|
||||
end
|
||||
|
||||
my_title='ELIMINATION OF THE MULTIPLIERS';
|
||||
lab1 = M_.endo_names(dr.order_var(dr.nstatic+[ 1 2 5:8]),:);
|
||||
labels = strvcat(lab1,lab1,M_.exo_names,M_.exo_names);
|
||||
lh = size(labels,2)+2;
|
||||
dyntable(my_title,headers,labels,res_table,lh,10,6);
|
||||
disp(' ')
|
||||
end
|
||||
|
|
|
|||
|
|
@ -0,0 +1,67 @@
|
|||
// Test of mult_elimination function
|
||||
|
||||
// parameters value is set to posterior mean as computed by ls1.mod
|
||||
// dR in objective function
|
||||
|
||||
var A de dq dR pie pie_obs pie_s R R_obs y y_obs y_s ;
|
||||
varexo e_A e_pies e_q e_ys ;
|
||||
|
||||
parameters psi1 psi2 psi3 rho_R tau alpha rr k rho_q rho_A rho_ys rho_pies ls1_r s1 s2 s3 s4 s5;
|
||||
|
||||
psi1 = 1.54;
|
||||
psi2 = 0.25;
|
||||
psi3 = 0.25;
|
||||
rho_R = 0.5;
|
||||
alpha = 0.3;
|
||||
rr = 2.51;
|
||||
k = 0.5;
|
||||
tau = 0.5;
|
||||
rho_q = 0.4;
|
||||
rho_A = 0.2;
|
||||
rho_ys = 0.9;
|
||||
rho_pies = 0.7;
|
||||
|
||||
model(linear);
|
||||
y = y(+1) - (tau +alpha*(2-alpha)*(1-tau))*(R-pie(+1))-alpha*(tau +alpha*(2-alpha)*(1-tau))*dq(+1) + alpha*(2-alpha)*((1-tau)/tau)*(y_s-y_s(+1))-A(+1);
|
||||
pie = exp(-rr/400)*pie(+1)+alpha*exp(-rr/400)*dq(+1)-alpha*dq+(k/(tau+alpha*(2-alpha)*(1-tau)))*y+alpha*(2-alpha)*(1-tau)/(tau*(tau+alpha*(2-alpha)*(1-tau)))*y_s;
|
||||
pie = de+(1-alpha)*dq+pie_s;
|
||||
//R = rho_R*R(-1)+(1-rho_R)*(psi1*pie+psi2*(y+alpha*(2-alpha)*((1-tau)/tau)*y_s)+psi3*de)+e_R;
|
||||
dq = rho_q*dq(-1)+e_q;
|
||||
y_s = rho_ys*y_s(-1)+e_ys;
|
||||
pie_s = rho_pies*pie_s(-1)+e_pies;
|
||||
A = rho_A*A(-1)+e_A;
|
||||
y_obs = y-y(-1)+A;
|
||||
pie_obs = 4*pie;
|
||||
R_obs = 4*R;
|
||||
dR = R-R(-1);
|
||||
end;
|
||||
|
||||
shocks;
|
||||
var e_q = 2.5^2;
|
||||
var e_A = 1.89;
|
||||
var e_ys = 1.89;
|
||||
var e_pies = 1.89;
|
||||
end;
|
||||
|
||||
planner_objective 0.25*pie_obs^2+y^2+0.1*dR^2;
|
||||
|
||||
ramsey_policy(order=1,irf=0,planner_discount=0.95);
|
||||
|
||||
dr2 = mult_elimination({'R'},M_,options_,oo_);
|
||||
|
||||
k1 = oo_.dr.nstatic+(1:oo_.dr.npred);
|
||||
k2 = strmatch('mult_',M_.endo_names(oo_.dr.order_var(k1),:));
|
||||
k3 = k1(setdiff(1:oo_.dr.npred,k2));
|
||||
k4 = oo_.dr.order_var(k3);
|
||||
|
||||
V0 = oo_.var(k4,k4);
|
||||
|
||||
|
||||
Atest = [dr2.M1(k3,:) dr2.M2(k3,:) dr2.M4(k3,:); eye(6) zeros(6,10);zeros(4,16)];
|
||||
Btest = [dr2.M3(k3,:); zeros(6,4); eye(4)];
|
||||
|
||||
V1=lyapunov_symm(Atest,Btest*M_.Sigma_e*Btest',options_.qz_criterium,options_.lyapunov_complex_threshold);
|
||||
|
||||
if max(max(abs(V1(1:6,1:6)-V0)))
|
||||
disp('Test OK')
|
||||
end
|
||||
Loading…
Reference in New Issue