v4: changes around restricted state space representation
git-svn-id: https://www.dynare.org/svn/dynare/dynare_v4@918 ac1d8469-bf42-47a9-8791-bf33cf982152time-shift
michel
parent
8fcac85ea1
commit
cb9fb1b353
|
|
@ -90,8 +90,9 @@ function [fval,cost_flag,ys,trend_coeff,info] = DsgeLikelihood(xparam1,gend,data
|
|||
%------------------------------------------------------------------------------
|
||||
% 2. call model setup & reduction program
|
||||
%------------------------------------------------------------------------------
|
||||
[T,R,SteadyState,info] = dynare_resolve;
|
||||
rs = bayestopt_.restrict_state;
|
||||
[T,R,SteadyState,info] = dynare_resolve(bayestopt_.restrict_var_list,...
|
||||
bayestopt_.restrict_columns,...
|
||||
bayestopt_.restrict_aux);
|
||||
if info(1) == 1 | info(1) == 2 | info(1) == 5
|
||||
fval = bayestopt_.penalty+1;
|
||||
cost_flag = 0;
|
||||
|
|
@ -101,8 +102,6 @@ function [fval,cost_flag,ys,trend_coeff,info] = DsgeLikelihood(xparam1,gend,data
|
|||
cost_flag = 0;
|
||||
return
|
||||
end
|
||||
T = T(rs,rs);
|
||||
R = R(rs,:);
|
||||
bayestopt_.mf = bayestopt_.mf1;
|
||||
if ~options_.noconstant
|
||||
if options_.loglinear == 1
|
||||
|
|
@ -129,21 +128,19 @@ function [fval,cost_flag,ys,trend_coeff,info] = DsgeLikelihood(xparam1,gend,data
|
|||
% 3. Initial condition of the Kalman filter
|
||||
%------------------------------------------------------------------------------
|
||||
if options_.lik_init == 1 % Kalman filter
|
||||
Pstar = lyapunov_symm(T,R*Q*transpose(R));
|
||||
Pstar = lyapunov_symm(T,R*Q*R');
|
||||
Pinf = [];
|
||||
elseif options_.lik_init == 2 % Old Diffuse Kalman filter
|
||||
Pstar = 10*eye(np);
|
||||
Pinf = [];
|
||||
elseif options_.lik_init == 3 % Diffuse Kalman filter
|
||||
Pstar = zeros(np,np);
|
||||
% ivs = bayestopt_.i_T_var_stable;
|
||||
% Pstar(ivs,ivs) = lyapunov_symm(T(ivs,ivs),R(ivs,:)*Q* ...
|
||||
% transpose(R(ivs,:)));
|
||||
Pstar(ivs,ivs) = lyapunov_symm(T(ivs,ivs),R(ivs,:)*Q* ...
|
||||
transpose(R(ivs,:)));
|
||||
Pinf = bayestopt_.Pinf;
|
||||
ivs = bayestopt_.restrict_var_list_stationary;
|
||||
R1 = R(ivs,:);
|
||||
Pstar(ivs,ivs) = lyapunov_symm(T(ivs,ivs),R1*Q*R1');
|
||||
% Pinf = bayestopt_.Pinf;
|
||||
% by M. Ratto
|
||||
% RR=T(:,find(~ismember([1:np],ivs)));
|
||||
RR=T(:,bayestopt_.restrict_var_list_nonstationary);
|
||||
i=find(abs(RR)>1.e-10);
|
||||
R0=zeros(size(RR));
|
||||
R0(i)=sign(RR(i));
|
||||
|
|
|
|||
|
|
@ -142,63 +142,15 @@ np = estim_params_.np ;
|
|||
nx = nvx+nvn+ncx+ncn+np;
|
||||
|
||||
dr = set_state_space([]);
|
||||
|
||||
%% Initialization with unit-root variables
|
||||
if ~isempty(options_.unit_root_vars)
|
||||
n_ur = size(options_.unit_root_vars,1);
|
||||
i_ur = zeros(n_ur,1);
|
||||
for i=1:n_ur
|
||||
i1 = strmatch(deblank(options_.unit_root_vars(i,:)),M_.endo_names(dr.order_var,:),'exact');
|
||||
if isempty(i1)
|
||||
error('Undeclared variable in unit_root_vars statement')
|
||||
end
|
||||
i_ur(i) = i1;
|
||||
end
|
||||
if M_.maximum_lag > 1
|
||||
l1 = flipud([cumsum(M_.lead_lag_incidence(1:M_.maximum_lag-1,dr.order_var),1);ones(1,M_.endo_nbr)]);
|
||||
n1 = nnz(l1);
|
||||
bayestopt_.Pinf = zeros(n1,n1);
|
||||
l2 = find(l1');
|
||||
l3 = zeros(M_.endo_nbr,M_.maximum_lag);
|
||||
l3(i_ur,:) = l1(:,i_ur)';
|
||||
l3 = l3(:);
|
||||
i_ur1 = find(l3(l2));
|
||||
i_stable = ones(M_.endo_nbr,1);
|
||||
i_stable(i_ur) = zeros(n_ur,1);
|
||||
i_stable = find(i_stable);
|
||||
bayestopt_.Pinf(i_ur1,i_ur1) = diag(ones(1,length(i_ur1)));
|
||||
bayestopt_.i_var_stable = i_stable;
|
||||
l3 = zeros(M_.endo_nbr,M_.maximum_lag);
|
||||
l3(i_stable,:) = l1(:,i_stable)';
|
||||
l3 = l3(:);
|
||||
bayestopt_.i_T_var_stable = find(l3(l2));
|
||||
else
|
||||
n1 = M_.endo_nbr;
|
||||
bayestopt_.Pinf = zeros(n1,n1);
|
||||
bayestopt_.Pinf(i_ur,i_ur) = diag(ones(1,length(i_ur)));
|
||||
l1 = ones(M_.endo_nbr,1);
|
||||
l1(i_ur,:) = zeros(length(i_ur),1);
|
||||
bayestopt_.i_T_var_stable = find(l1);
|
||||
end
|
||||
options_.lik_init = 3;
|
||||
end % if ~isempty(options_.unit_root_vars)
|
||||
|
||||
if isempty(options_.datafile)
|
||||
error('ESTIMATION: datafile option is missing')
|
||||
end
|
||||
nstatic = dr.nstatic;
|
||||
npred = dr.npred;
|
||||
nspred = dr.nspred;
|
||||
|
||||
if isempty(options_.varobs)
|
||||
error('ESTIMATION: VAROBS is missing')
|
||||
end
|
||||
|
||||
|
||||
%% If jscale isn't specified for an estimated parameter, use
|
||||
%% global option options_.jscale, set to 0.2, by default
|
||||
k = find(isnan(bayestopt_.jscale));
|
||||
bayestopt_.jscale(k) = options_.mh_jscale;
|
||||
|
||||
%% Read and demean data
|
||||
rawdata = read_variables(options_.datafile,options_.varobs,[],options_.xls_sheet,options_.xls_range);
|
||||
%% Setting resticted state space (observed + predetermined variables)
|
||||
|
||||
k = [];
|
||||
k1 = [];
|
||||
|
|
@ -213,7 +165,7 @@ k2 = [k2;[M_.endo_nbr+(1:dr.nspred-dr.npred)]'];
|
|||
|
||||
% set restrict_state to postion of observed + state variables
|
||||
% in expanded state vector
|
||||
bayestopt_.restrict_state = k2;
|
||||
bayestopt_.restrict_var_list = k2;
|
||||
% set mf1 to positions of observed variables in restricted state vector
|
||||
% for likelihood computation
|
||||
[junk,bayestopt_.mf1] = ismember(k,k2);
|
||||
|
|
@ -221,6 +173,55 @@ bayestopt_.restrict_state = k2;
|
|||
% for filtering and smoothing
|
||||
bayestopt_.mf2 = k;
|
||||
bayestopt_.mfys = k1;
|
||||
|
||||
[junk,ic] = intersect(k2,nstatic+(1:npred)');
|
||||
bayestopt_.restrict_columns = [ic length(k2)+(1:nspred-npred)]';
|
||||
aux = dr.transition_auxiliary_variables;
|
||||
k = find(aux(:,2) > npred);
|
||||
aux(:,2) = aux(:,2)+sum(k2 <= nstatic);
|
||||
bayestopt_.restrict_aux = aux;
|
||||
|
||||
|
||||
%% Initialization with unit-root variables
|
||||
if ~isempty(options_.unit_root_vars)
|
||||
n_ur = size(options_.unit_root_vars,1);
|
||||
i_ur = zeros(n_ur,1);
|
||||
for i=1:n_ur
|
||||
i1 = strmatch(deblank(options_.unit_root_vars(i,:)),M_.endo_names(dr.order_var,:),'exact');
|
||||
if isempty(i1)
|
||||
error('Undeclared variable in unit_root_vars statement')
|
||||
end
|
||||
i_ur(i) = i1;
|
||||
end
|
||||
[junk,bayestopt_.restrict_var_list_stationary] = ...
|
||||
setdiff(bayestopt_.restrict_var_list,i_ur);
|
||||
[junk,bayestopt_.restrict_var_list_nonstationary] = ...
|
||||
setdiff(bayestopt_.restrict_var_list,i_ur);
|
||||
if M_.maximum_lag > 1
|
||||
l1 = flipud([cumsum(M_.lead_lag_incidence(1:M_.maximum_lag-1,dr.order_var),1);ones(1,M_.endo_nbr)]);
|
||||
l2 = l1(:,restrict_var_list);
|
||||
il2 = find(l2' > 0);
|
||||
l2(il2) = (1:length(il2))';
|
||||
bayestopt_.restict_var_list_stationary = ...
|
||||
nonzeros(l2(:,restrict_var_list_stationary));
|
||||
bayestopt_.restict_var_list_nonstationary = ...
|
||||
nonzeros(l2(:,restrict_var_list_nonstationary));
|
||||
end
|
||||
options_.lik_init = 3;
|
||||
end % if ~isempty(options_.unit_root_vars)
|
||||
|
||||
if isempty(options_.datafile)
|
||||
error('ESTIMATION: datafile option is missing')
|
||||
end
|
||||
|
||||
%% If jscale isn't specified for an estimated parameter, use
|
||||
%% global option options_.jscale, set to 0.2, by default
|
||||
k = find(isnan(bayestopt_.jscale));
|
||||
bayestopt_.jscale(k) = options_.mh_jscale;
|
||||
|
||||
%% Read and demean data
|
||||
rawdata = read_variables(options_.datafile,options_.varobs,[],options_.xls_sheet,options_.xls_range);
|
||||
|
||||
options_ = set_default_option(options_,'nobs',size(rawdata,1)-options_.first_obs+1);
|
||||
gend = options_.nobs;
|
||||
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
function [A,B,ys,info] = dynare_resolve()
|
||||
global oo_
|
||||
function [A,B,ys,info] = dynare_resolve(iv,ic,aux)
|
||||
global oo_ M_
|
||||
|
||||
[oo_.dr,info] = resol(oo_.steady_state,0);
|
||||
|
||||
|
|
@ -10,5 +10,15 @@ function [A,B,ys,info] = dynare_resolve()
|
|||
return
|
||||
end
|
||||
|
||||
[A,B] = kalman_transition_matrix(oo_.dr);
|
||||
if nargin == 0
|
||||
endo_nbr = M_.endo_nbr;
|
||||
iv = (1:endo_nbr)';
|
||||
ic = [ nstatic+(1:npred) endo_nbr+(1:size(dr.ghx,2)-npred) ]';
|
||||
aux = oo_.dr.transtion_auxiliary_variables(:,2);
|
||||
k = find(aux > oo_.dr.npred);
|
||||
aux = aux + nstatic;
|
||||
aux(k) = aux(k) + oo_.dr.nfrwd;
|
||||
end
|
||||
|
||||
[A,B] = kalman_transition_matrix(oo_.dr,iv,ic,aux);
|
||||
ys = oo_.dr.ys;
|
||||
|
|
@ -1,4 +1,4 @@
|
|||
function [loss,vx,info]=osr_obj(x,i_params,weights);
|
||||
function [loss,vx,info]=osr_obj(x,i_params,i_var,weights);
|
||||
% objective function for optimal simple rules (OSR)
|
||||
global M_ oo_ optimal_Q_ it_
|
||||
% global ys_ Sigma_e_ endo_nbr exo_nbr optimal_Q_ it_ ykmin_ options_
|
||||
|
|
@ -33,16 +33,7 @@ function [loss,vx,info]=osr_obj(x,i_params,weights);
|
|||
otherwise
|
||||
end
|
||||
|
||||
[A,B] = kalman_transition_matrix(dr);
|
||||
[vx,ns_var] = lyapunov_symm(A,B*M_.Sigma_e*B');
|
||||
endo_nbr = M_.endo_nbr;
|
||||
i_var = (1:endo_nbr)';
|
||||
i_var(ns_var) = zeros(length(ns_var),1);
|
||||
i_var = nonzeros(i_var);
|
||||
vx = vx(i_var,i_var);
|
||||
weights = weights(dr.order_var,dr.order_var);
|
||||
weights = sparse(weights(i_var,i_var));
|
||||
|
||||
vx = get_variance_of_endogenous_variables(dr,i_var);
|
||||
loss = weights(:)'*vx(:);
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -28,7 +28,7 @@ npred = length(pred_var);
|
|||
nfwrd = length(fwrd_var);
|
||||
nstatic = length(stat_var);
|
||||
order_var = [ stat_var; pred_var; both_var; fwrd_var];
|
||||
inv_order_var(order_var) = (1:M_.endo_nbr)';
|
||||
inv_order_var(order_var) = (1:M_.endo_nbr);
|
||||
|
||||
% building kmask for z state vector in t+1
|
||||
if M_.maximum_lag > 0
|
||||
|
|
@ -72,7 +72,7 @@ kstate(M_.maximum_lead*M_.endo_nbr+1:end,4) = kiy((M_.maximum_lead+1)*M_.endo_nb
|
|||
kstate = kstate(i_kmask,:);
|
||||
|
||||
dr.order_var = order_var;
|
||||
dr.inv_order_var = inv_order_var;
|
||||
dr.inv_order_var = inv_order_var';
|
||||
dr.nstatic = nstatic;
|
||||
dr.npred = npred+nboth;
|
||||
dr.kstate = kstate;
|
||||
|
|
@ -85,9 +85,10 @@ dr.nsfwrd = sum(kstate(:,2) > M_.maximum_lag+1);
|
|||
% number of predetermined variables in the state vector
|
||||
dr.nspred = sum(kstate(:,2) <= M_.maximum_lag+1);
|
||||
|
||||
% copmutes column position of auxiliary variables for
|
||||
% compact transition matrix (only state variables)
|
||||
aux = zeros(0,2);
|
||||
k0 = kstate(find(kstate(:,2) <= M_.maximum_lag+1),:);;
|
||||
offset_col = nstatic;
|
||||
for i=M_.maximum_lag:-1:2
|
||||
i1 = find(k0(:,2) == i);
|
||||
n1 = size(i1,1);
|
||||
|
|
@ -95,9 +96,6 @@ for i=M_.maximum_lag:-1:2
|
|||
for j1 = 1:n1
|
||||
j(j1) = find(k0(i0,1)==k0(i1(j1),1));
|
||||
end
|
||||
if i == M_.maximum_lag-1
|
||||
offset_col = dr.nstatic+dr.nfwrd;
|
||||
end
|
||||
aux = [aux; offset_col+i0(j)];
|
||||
i0 = i1;
|
||||
end
|
||||
|
|
|
|||
Loading…
Reference in New Issue