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

Fixed indentation of matlab files.

time-shift
Stéphane Adjemian (Charybdis) 2017-05-16 15:10:20 +02:00
parent 3d91cf9716
commit 5417b27ac7
566 changed files with 18524 additions and 18688 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

185
matlab/AHessian.m
View File

@ -26,107 +26,107 @@ function [AHess, DLIK, LIK] = AHessian(T,R,Q,H,P,Y,DT,DYss,DOm,DH,DP,start,mf,ka
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
k = size(DT,3); % number of structural parameters
smpl = size(Y,2); % Sample size.
pp = size(Y,1); % Maximum number of observed variables.
mm = size(T,2); % Number of state variables.
a = zeros(mm,1); % State vector.
Om = R*Q*transpose(R); % Variance of R times the vector of structural innovations.
t = 0; % Initialization of the time index.
oldK = 0;
notsteady = 1; % Steady state flag.
F_singular = 1;
k = size(DT,3); % number of structural parameters
smpl = size(Y,2); % Sample size.
pp = size(Y,1); % Maximum number of observed variables.
mm = size(T,2); % Number of state variables.
a = zeros(mm,1); % State vector.
Om = R*Q*transpose(R); % Variance of R times the vector of structural innovations.
t = 0; % Initialization of the time index.
oldK = 0;
notsteady = 1; % Steady state flag.
F_singular = 1;
lik = zeros(smpl,1); % Initialization of the vector gathering the densities.
LIK = Inf; % Default value of the log likelihood.
if nargout > 1
DLIK = zeros(k,1); % Initialization of the score.
end
AHess = zeros(k,k); % Initialization of the Hessian
Da = zeros(mm,k); % State vector.
Dv = zeros(length(mf),k);
AHess = zeros(k,k); % Initialization of the Hessian
Da = zeros(mm,k); % State vector.
Dv = zeros(length(mf),k);
% for ii = 1:k
% DOm = DR(:,:,ii)*Q*transpose(R) + R*DQ(:,:,ii)*transpose(R) + R*Q*transpose(DR(:,:,ii));
% DOm = DR(:,:,ii)*Q*transpose(R) + R*DQ(:,:,ii)*transpose(R) + R*Q*transpose(DR(:,:,ii));
% end
while notsteady && t<smpl
t = t+1;
v = Y(:,t)-a(mf);
F = P(mf,mf) + H;
if rcond(F) < kalman_tol
if ~all(abs(F(:))<kalman_tol)
return
else
a = T*a;
P = T*P*transpose(T)+Om;
end
while notsteady && t<smpl
t = t+1;
v = Y(:,t)-a(mf);
F = P(mf,mf) + H;
if rcond(F) < kalman_tol
if ~all(abs(F(:))<kalman_tol)
return
else
F_singular = 0;
iF = inv(F);
K = P(:,mf)*iF;
lik(t) = log(det(F))+transpose(v)*iF*v;
[DK,DF,DP1] = computeDKalman(T,DT,DOm,P,DP,DH,mf,iF,K);
for ii = 1:k
Dv(:,ii) = -Da(mf,ii) - DYss(mf,ii);
Da(:,ii) = DT(:,:,ii)*(a+K*v) + T*(Da(:,ii)+DK(:,:,ii)*v + K*Dv(:,ii));
if t>=start && nargout > 1
DLIK(ii,1) = DLIK(ii,1) + trace( iF*DF(:,:,ii) ) + 2*Dv(:,ii)'*iF*v - v'*(iF*DF(:,:,ii)*iF)*v;
end
end
vecDPmf = reshape(DP(mf,mf,:),[],k);
% iPmf = inv(P(mf,mf));
if t>=start
AHess = AHess + Dv'*iF*Dv + .5*(vecDPmf' * kron(iF,iF) * vecDPmf);
end
a = T*(a+K*v);
P = T*(P-K*P(mf,:))*transpose(T)+Om;
DP = DP1;
a = T*a;
P = T*P*transpose(T)+Om;
end
notsteady = max(max(abs(K-oldK))) > riccati_tol;
oldK = K;
end
else
F_singular = 0;
iF = inv(F);
K = P(:,mf)*iF;
lik(t) = log(det(F))+transpose(v)*iF*v;
if F_singular
error('The variance of the forecast error remains singular until the end of the sample')
end
[DK,DF,DP1] = computeDKalman(T,DT,DOm,P,DP,DH,mf,iF,K);
if t < smpl
t0 = t+1;
while t < smpl
t = t+1;
v = Y(:,t)-a(mf);
for ii = 1:k
Dv(:,ii) = -Da(mf,ii)-DYss(mf,ii);
Da(:,ii) = DT(:,:,ii)*(a+K*v) + T*(Da(:,ii)+DK(:,:,ii)*v + K*Dv(:,ii));
if t>=start && nargout >1
DLIK(ii,1) = DLIK(ii,1) + trace( iF*DF(:,:,ii) ) + 2*Dv(:,ii)'*iF*v - v'*(iF*DF(:,:,ii)*iF)*v;
end
end
if t>=start
AHess = AHess + Dv'*iF*Dv;
end
a = T*(a+K*v);
lik(t) = transpose(v)*iF*v;
end
AHess = AHess + .5*(smpl-t0+1)*(vecDPmf' * kron(iF,iF) * vecDPmf);
if nargout > 1
for ii = 1:k
% DLIK(ii,1) = DLIK(ii,1) + (smpl-t0+1)*trace( iF*DF(:,:,ii) );
Dv(:,ii) = -Da(mf,ii) - DYss(mf,ii);
Da(:,ii) = DT(:,:,ii)*(a+K*v) + T*(Da(:,ii)+DK(:,:,ii)*v + K*Dv(:,ii));
if t>=start && nargout > 1
DLIK(ii,1) = DLIK(ii,1) + trace( iF*DF(:,:,ii) ) + 2*Dv(:,ii)'*iF*v - v'*(iF*DF(:,:,ii)*iF)*v;
end
end
vecDPmf = reshape(DP(mf,mf,:),[],k);
% iPmf = inv(P(mf,mf));
if t>=start
AHess = AHess + Dv'*iF*Dv + .5*(vecDPmf' * kron(iF,iF) * vecDPmf);
end
lik(t0:smpl) = lik(t0:smpl) + log(det(F));
% for ii = 1:k;
% for jj = 1:ii
% H(ii,jj) = trace(iPmf*(.5*DP(mf,mf,ii)*iPmf*DP(mf,mf,jj) + Dv(:,ii)*Dv(:,jj)'));
% end
% end
end
AHess = -AHess;
a = T*(a+K*v);
P = T*(P-K*P(mf,:))*transpose(T)+Om;
DP = DP1;
end
notsteady = max(max(abs(K-oldK))) > riccati_tol;
oldK = K;
end
if F_singular
error('The variance of the forecast error remains singular until the end of the sample')
end
if t < smpl
t0 = t+1;
while t < smpl
t = t+1;
v = Y(:,t)-a(mf);
for ii = 1:k
Dv(:,ii) = -Da(mf,ii)-DYss(mf,ii);
Da(:,ii) = DT(:,:,ii)*(a+K*v) + T*(Da(:,ii)+DK(:,:,ii)*v + K*Dv(:,ii));
if t>=start && nargout >1
DLIK(ii,1) = DLIK(ii,1) + trace( iF*DF(:,:,ii) ) + 2*Dv(:,ii)'*iF*v - v'*(iF*DF(:,:,ii)*iF)*v;
end
end
if t>=start
AHess = AHess + Dv'*iF*Dv;
end
a = T*(a+K*v);
lik(t) = transpose(v)*iF*v;
end
AHess = AHess + .5*(smpl-t0+1)*(vecDPmf' * kron(iF,iF) * vecDPmf);
if nargout > 1
for ii = 1:k
% DLIK(ii,1) = DLIK(ii,1) + (smpl-t0+1)*trace( iF*DF(:,:,ii) );
end
end
lik(t0:smpl) = lik(t0:smpl) + log(det(F));
% for ii = 1:k;
% for jj = 1:ii
% H(ii,jj) = trace(iPmf*(.5*DP(mf,mf,ii)*iPmf*DP(mf,mf,jj) + Dv(:,ii)*Dv(:,jj)'));
% end
% end
end
AHess = -AHess;
if nargout > 1
DLIK = DLIK/2;
end
@ -134,15 +134,15 @@ end
lik = (lik + pp*log(2*pi))/2;
LIK = sum(lik(start:end)); % Minus the log-likelihood.
% end of main function
% end of main function
function [DK,DF,DP1] = computeDKalman(T,DT,DOm,P,DP,DH,mf,iF,K)
k = size(DT,3);
tmp = P-K*P(mf,:);
k = size(DT,3);
tmp = P-K*P(mf,:);
for ii = 1:k
DF(:,:,ii) = DP(mf,mf,ii) + DH(:,:,ii);
DF(:,:,ii) = DP(mf,mf,ii) + DH(:,:,ii);
DiF(:,:,ii) = -iF*DF(:,:,ii)*iF;
DK(:,:,ii) = DP(:,mf,ii)*iF + P(:,mf)*DiF(:,:,ii);
Dtmp = DP(:,:,ii) - DK(:,:,ii)*P(mf,:) - K*DP(mf,:,ii);
@ -150,6 +150,3 @@ for ii = 1:k
end
% end of computeDKalman

2
matlab/AIM/SPAimerr.m
View File

@ -29,7 +29,7 @@ function e = SPAimerr(c)
% Journal of Economic Dynamics and Control, 2010, vol. 34, issue 3,
% pages 472-489
if(c==1) e='Aim: unique solution.';
if(c==1) e='Aim: unique solution.';
elseif(c==2) e='Aim: roots not correctly computed by real_schur.';
elseif(c==3) e='Aim: too many big roots.';
elseif(c==35) e='Aim: too many big roots, and q(:,right) is singular.';

24
matlab/AIM/SPAmalg.m
View File

@ -1,5 +1,5 @@
function [b,rts,ia,nexact,nnumeric,lgroots,aimcode] = ...
SPAmalg(h,neq,nlag,nlead,condn,uprbnd)
SPAmalg(h,neq,nlag,nlead,condn,uprbnd)
% [b,rts,ia,nexact,nnumeric,lgroots,aimcode] = ...
% SPAmalg(h,neq,nlag,nlead,condn,uprbnd)
%
@ -8,9 +8,9 @@ function [b,rts,ia,nexact,nnumeric,lgroots,aimcode] = ...
% roots. This procedure will fail if the companion matrix is
% defective and does not have a linearly independent set of
% eigenvectors associated with the big roots.
%
%
% Input arguments:
%
%
% h Structural coefficient matrix (neq,neq*(nlag+1+nlead)).
% neq Number of equations.
% nlag Number of lags.
@ -19,9 +19,9 @@ function [b,rts,ia,nexact,nnumeric,lgroots,aimcode] = ...
% by numeric_shift and reduced_form.
% uprbnd Inclusive upper bound for the modulus of roots
% allowed in the reduced form.
%
%
% Output arguments:
%
%
% b Reduced form coefficient matrix (neq,neq*nlag).
% rts Roots returned by eig.
% ia Dimension of companion matrix (number of non-trivial
@ -57,7 +57,7 @@ function [b,rts,ia,nexact,nnumeric,lgroots,aimcode] = ...
% pages 472-489
b=[];rts=[];ia=[];nexact=[];nnumeric=[];lgroots=[];aimcode=[];
if(nlag<1 || nlead<1)
if(nlag<1 || nlead<1)
error('Aim_eig: model must have at least one lag and one lead');
end
% Initialization.
@ -75,17 +75,17 @@ if (iq>qrows)
end
[a,ia,js] = SPBuild_a(h,qcols,neq);
if (ia ~= 0)
if any(any(isnan(a))) || any(any(isinf(a)))
if any(any(isnan(a))) || any(any(isinf(a)))
disp('A is NAN or INF')
aimcode=63;
return
end
aimcode=63;
return
end
[w,rts,lgroots,flag_trouble]=SPEigensystem(a,uprbnd,min(length(js),qrows-iq+1));
if flag_trouble==1
disp('Problem in SPEIG');
disp('Problem in SPEIG');
aimcode=64;
return
end
end
q = SPCopy_w(q,w,js,iq,qrows);
end
test=nexact+nnumeric+lgroots;

16
matlab/AIM/SPBuild_a.m
View File

@ -41,9 +41,9 @@ hs(:,left) = -hs(:,right)\hs(:,left);
a = zeros(qcols,qcols);
if(qcols > neq)
eyerows = 1:qcols-neq;
eyecols = neq+1:qcols;
a(eyerows,eyecols) = eye(qcols-neq);
eyerows = 1:qcols-neq;
eyecols = neq+1:qcols;
a(eyerows,eyecols) = eye(qcols-neq);
end
hrows = qcols-neq+1:qcols;
a(hrows,:) = hs(:,left);
@ -51,14 +51,14 @@ a(hrows,:) = hs(:,left);
% Delete inessential lags and build index array js. js indexes the
% columns in the big transition matrix that correspond to the
% essential lags in the model. They are the columns of q that will
% get the unstable left eigenvectors.
% get the unstable left eigenvectors.
js = 1:qcols;
zerocols = sum(abs(a)) == 0;
while( any(zerocols) )
a(:,zerocols) = [];
a(zerocols,:) = [];
js(zerocols) = [];
zerocols = sum(abs(a)) == 0;
a(:,zerocols) = [];
a(zerocols,:) = [];
js(zerocols) = [];
zerocols = sum(abs(a)) == 0;
end
ia = length(js);

8
matlab/AIM/SPCopy_w.m
View File

@ -2,7 +2,7 @@ function q = SPCopy_w(q,w,js,iq,qrows)
% q = SPCopy_w(q,w,js,iq,qrows)
%
% Copy the eigenvectors corresponding to the largest roots into the
% remaining empty rows and columns js of q
% remaining empty rows and columns js of q
% Author: Gary Anderson
% Original file downloaded from:
@ -30,7 +30,7 @@ function q = SPCopy_w(q,w,js,iq,qrows)
if(iq < qrows)
lastrows = iq+1:qrows;
wrows = 1:length(lastrows);
q(lastrows,js) = w(:,wrows)';
lastrows = iq+1:qrows;
wrows = 1:length(lastrows);
q(lastrows,js) = w(:,wrows)';
end

38
matlab/AIM/SPEigensystem.m
View File

@ -30,7 +30,7 @@ function [w,rts,lgroots,flag_trouble] = SPEigensystem(a,uprbnd,rowsLeft)
% Journal of Economic Dynamics and Control, 2010, vol. 34, issue 3,
% pages 472-489
opts.disp=0;
opts.disp=0;
% next block is commented out because eigs() intermitently returns different rts
%try
% [w,d] = eigs(a',rowsLeft,'LM',opts);
@ -39,24 +39,24 @@ opts.disp=0;
% [mag,k] = sort(-mag);
% rts = rts(k);
%catch
%disp('Catch in SPE');
%pause(0.5);
%aStr=datestr(clock);
%eval(['save ' regexprep(aStr,' ','') ' a']);
try
[w,d]=eig(a');
catch
lasterr
w=[];rts=[];lgroots=[];
flag_trouble=1;
return
end
rts = diag(d);
mag = abs(rts);
[mag,k] = sort(-mag);
rts = rts(k);
%disp('Catch in SPE');
%pause(0.5);
%aStr=datestr(clock);
%eval(['save ' regexprep(aStr,' ','') ' a']);
try
[w,d]=eig(a');
catch
lasterr
w=[];rts=[];lgroots=[];
flag_trouble=1;
return
end
rts = diag(d);
mag = abs(rts);
[mag,k] = sort(-mag);
rts = rts(k);
%end
flag_trouble=0;
flag_trouble=0;
%ws=SPSparse(w);
ws=sparse(w);
@ -65,7 +65,7 @@ ws = ws(:,k);
% Given a complex conjugate pair of vectors W = [w1,w2], there is a
% nonsingular matrix D such that W*D = real(W) + imag(W). That is to
% say, W and real(W)+imag(W) span the same subspace, which is all
% that aim cares about.
% that aim cares about.
ws = real(ws) + imag(ws);

13
matlab/AIM/SPExact_shift.m
View File

@ -36,12 +36,11 @@ right = qcols+1:qcols+neq;
zerorows = find( sum(abs( hs(:,right)' ))==0 );
while( any(zerorows) && iq <= qrows )
nz = length(zerorows);
q(iq+1:iq+nz,:) = hs(zerorows,left);
hs(zerorows,:) = SPShiftright(hs(zerorows,:),neq);
iq = iq + nz;
nexact = nexact + nz;
zerorows = find( sum(abs( hs(:,right)' ))==0 );
nz = length(zerorows);
q(iq+1:iq+nz,:) = hs(zerorows,left);
hs(zerorows,:) = SPShiftright(hs(zerorows,:),neq);
iq = iq + nz;
nexact = nexact + nz;
zerorows = find( sum(abs( hs(:,right)' ))==0 );
end
h=full(hs);

20
matlab/AIM/SPNumeric_shift.m
View File

@ -37,14 +37,14 @@ right = qcols+1:qcols+neq;
zerorows = find( abs(diag(R)) <= condn );
while( any(zerorows) && iq <= qrows )
h=sparse(h);
Q=sparse(Q);
h = Q'*h;
nz = length(zerorows);
q(iq+1:iq+nz,:) = h(zerorows,left);
h(zerorows,:) = SPShiftright( h(zerorows,:), neq );
iq = iq + nz;
nnumeric = nnumeric + nz;
[Q,R,E] = qr( full(h(:,right)) );
zerorows = find( abs(diag(R)) <= condn );
h=sparse(h);
Q=sparse(Q);
h = Q'*h;
nz = length(zerorows);
q(iq+1:iq+nz,:) = h(zerorows,left);
h(zerorows,:) = SPShiftright( h(zerorows,:), neq );
iq = iq + nz;
nnumeric = nnumeric + nz;
[Q,R,E] = qr( full(h(:,right)) );
zerorows = find( abs(diag(R)) <= condn );
end

14
matlab/AIM/SPObstruct.m
View File

@ -2,7 +2,7 @@ function scof = SPObstruct(cof,cofb,neq,nlag,nlead)
% scof = SPObstruct(cof,cofb,neq,nlag,nlead)
%
% Construct the coefficients in the observable structure.
%
%
% Input arguments:
%
% cof structural coefficients
@ -51,17 +51,17 @@ qs=sparse(q);
qs(1:rc,1:cc) = sparse(cofb);
qcols = neq*(nlag+nlead);
if( nlead > 1 )
for i = 1:nlead-1
rows = i*neq + (1:neq);
qs(rows,:) = SPShiftright( qs((rows-neq),:), neq );
end
if( nlead > 1 )
for i = 1:nlead-1
rows = i*neq + (1:neq);
qs(rows,:) = SPShiftright( qs((rows-neq),:), neq );
end
end
l = (1: neq*nlag);
r = (neq*nlag+1: neq*(nlag+nlead));
qs(:,l) = - qs(:,r) \ qs(:,l);
qs(:,l) = - qs(:,r) \ qs(:,l);
minus = 1: neq*(nlag+1);
plus = neq*(nlag+1)+1: neq*(nlag+1+nlead);

3
matlab/AIM/SPReduced_form.m
View File

@ -38,7 +38,7 @@ if(nonsing)
b = qs(1:neq,1:bcols);
b = full(b);
else %rescale by dividing row by maximal qr element
%'inverse condition number small, rescaling '
%'inverse condition number small, rescaling '
themax=max(abs(qs(:,right)),[],2);
oneover = diag(1 ./ themax);
nonsing = rcond(full(oneover *qs(:,right))) > condn;
@ -48,4 +48,3 @@ else %rescale by dividing row by maximal qr element
b = full(b);
end
end

2
matlab/AIM/SPShiftright.m
View File

@ -3,7 +3,7 @@ function [y] = SPShiftright(x,n)
% [y] = shiftright(x,n)
%
% Shift the rows of x to the right by n columns, leaving zeros in the
% first n columns.
% first n columns.
% Original author: Gary Anderson
% Original file downloaded from:

48
matlab/AIM/dynAIMsolver1.m
View File

@ -1,20 +1,20 @@
function [dr,aimcode,rts]=dynAIMsolver1(jacobia_,M_,dr)
% function [dr,aimcode]=dynAIMsolver1(jacobia_,M_,dr)
% Maps Dynare jacobia to AIM 1st order model solver designed and developed by Gary ANderson
% Maps Dynare jacobia to AIM 1st order model solver designed and developed by Gary ANderson
% and derives the solution for gy=dr.hgx and gu=dr.hgu from the AIM outputs
% AIM System is given as a sum:
% i.e. for i=-$...+& SUM(Hi*xt+i)= £*zt, t = 0, . . . ,?
% AIM System is given as a sum:
% i.e. for i=-$...+& SUM(Hi*xt+i)= £*zt, t = 0, . . . ,?
% and its input as single array of matrices: [H-$... Hi ... H+&]
% and its solution as xt=SUM( Bi*xt+i) + @*£*zt for i=-$...-1
% with the output in form bb=[B-$... Bi ... B-1] and @=inv(Ho+H1*B-1)
% Dynare jacobian = [fy'-$... fy'i ... fy'+& fu']
% where [fy'-$... fy'i ... fy'+&]=[H-$... Hi ... H+&] and fu'= £
% and its solution as xt=SUM( Bi*xt+i) + @*£*zt for i=-$...-1
% with the output in form bb=[B-$... Bi ... B-1] and @=inv(Ho+H1*B-1)
% Dynare jacobian = [fy'-$... fy'i ... fy'+& fu']
% where [fy'-$... fy'i ... fy'+&]=[H-$... Hi ... H+&] and fu'= £
%
% INPUTS
% jacobia_ [matrix] 1st order derivative of the model
% jacobia_ [matrix] 1st order derivative of the model
% dr [matlab structure] Decision rules for stochastic simulations.
% M_ [matlab structure] Definition of the model.
%
% M_ [matlab structure] Definition of the model.
%
% OUTPUTS
% dr [matlab structure] Decision rules for stochastic simulations.
% aimcode [integer] 1: the model defines variables uniquely
@ -31,20 +31,20 @@ function [dr,aimcode,rts]=dynAIMsolver1(jacobia_,M_,dr)
% (c==63) e='Aim: A is NAN or INF.';
% (c==64) e='Aim: Problem in SPEIG.';
% else e='Aimerr: return code not properly specified';
%
%
% SPECIAL REQUIREMENTS
% Dynare use:
% Dynare use:
% 1) the lognormal block in DR1 is being invoked for some models and changing
% values of ghx and ghy. We need to return the AIM output
% values before that block and run the block with the current returned values
% of gy (i.e. dr.ghx) and gu (dr.ghu) if it is needed even when the AIM is used
% of gy (i.e. dr.ghx) and gu (dr.ghu) if it is needed even when the AIM is used
% (it does not depend on mjdgges output).
%
% 2) passing in aa={Q'|1}*jacobia_ can produce ~ one order closer
% results to the Dynare solutiion then when if plain jacobia_ is passed,
% i.e. diff < e-14 for aa and diff < *e-13 for jacobia_ if Q' is used.
%
% GP July 2008
% 2) passing in aa={Q'|1}*jacobia_ can produce ~ one order closer
% results to the Dynare solutiion then when if plain jacobia_ is passed,
% i.e. diff < e-14 for aa and diff < *e-13 for jacobia_ if Q' is used.
%
% GP July 2008
% Copyright (C) 2008-2012 Dynare Team
%
@ -69,8 +69,8 @@ lags=M_.maximum_endo_lag; % no of lags and leads
leads=M_.maximum_endo_lead;
klen=(leads+lags+1); % total lenght
theAIM_H=zeros(neq, neq*klen); % alloc space
lli=M_.lead_lag_incidence';
% "sparse" the compact jacobia into AIM H aray of matrices
lli=M_.lead_lag_incidence';
% "sparse" the compact jacobia into AIM H aray of matrices
% without exogenous shoc
theAIM_H(:,find(lli(:)))=jacobia_(:,nonzeros(lli(:)));
condn = 1.e-10;%SPAmalg uses this in zero tests
@ -102,7 +102,7 @@ if aimcode==1 %if OK
col_order=[((i-1)*neq)+dr.order_var' col_order];
end
bb_ord= bb(dr.order_var,col_order); % derive ordered gy
% variables are present in the state space at the lag at which they
% appear and at all smaller lags. The are ordered from smaller to
% higher lag (reversed order of M_.lead_lag_incidence rows for lagged
@ -117,7 +117,7 @@ if aimcode==1 %if OK
%theH0= theAIM_H (:,lags*neq+1: (lags+1)*neq);
% theHP= theAIM_H (:,(M_.maximum_endo_lag+1)*neq+1: (M_.maximum_endo_lag+2)*neq);
%theHP= theAIM_H (:,(lags+1)*neq+1: (lags+2)*neq);
theAIM_Psi= - jacobia_(:, size(nonzeros(lli(:)))+1:end);%
theAIM_Psi= - jacobia_(:, size(nonzeros(lli(:)))+1:end);%
%? = inv(H0 + H1B1)
%phi= (theH0+theHP*sparse(bb(:,(lags-1)*neq+1:end)))\eye( neq);
%AIM_ghu=phi*theAIM_Psi;
@ -137,8 +137,8 @@ else
if aimcode < 1 || aimcode > 5 % too big exception, use mjdgges
error('Error in AIM: aimcode=%d ; %s', aimcode, err);
end
% if aimcode > 5
% if aimcode > 5
% disp(['Error in AIM: aimcode=' sprintf('%d : %s',aimcode, err)]);
% aimcode=5;
% end
% end
end

59
matlab/AIM_first_order_solver.m
View File

@ -67,35 +67,34 @@ function [dr,info]=AIM_first_order_solver(jacobia,M,dr,qz_criterium)
%
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
info = 0;
[dr,aimcode]=dynAIMsolver1(jacobia,M,dr);
if aimcode ~=1
info(1) = convertAimCodeToInfo(aimCode); %convert to be in the 100 range
info(2) = 1.0e+8;
return
info = 0;
[dr,aimcode]=dynAIMsolver1(jacobia,M,dr);
if aimcode ~=1
info(1) = convertAimCodeToInfo(aimCode); %convert to be in the 100 range
info(2) = 1.0e+8;
return
end
A = kalman_transition_matrix(dr,M.nstatic+(1:M.nspred), 1:M.nspred,...
M.exo_nbr);
dr.eigval = eig(A);
disp(dr.eigval)
nd = size(dr.kstate,1);
nba = nd-sum( abs(dr.eigval) < qz_criterium );
nsfwrd = M.nsfwrd;
if nba ~= nsfwrd
temp = sort(abs(dr.eigval));
if nba > nsfwrd
temp = temp(nd-nba+1:nd-nsfwrd)-1-qz_criterium;
info(1) = 3;
elseif nba < nsfwrd
temp = temp(nd-nsfwrd+1:nd-nba)-1-qz_criterium;
info(1) = 4;
end
A = kalman_transition_matrix(dr,M.nstatic+(1:M.nspred), 1:M.nspred,...
M.exo_nbr);
dr.eigval = eig(A);
disp(dr.eigval)
nd = size(dr.kstate,1);
nba = nd-sum( abs(dr.eigval) < qz_criterium );
nsfwrd = M.nsfwrd;
if nba ~= nsfwrd
temp = sort(abs(dr.eigval));
if nba > nsfwrd
temp = temp(nd-nba+1:nd-nsfwrd)-1-qz_criterium;
info(1) = 3;
elseif nba < nsfwrd
temp = temp(nd-nsfwrd+1:nd-nba)-1-qz_criterium;
info(1) = 4;
end
info(2) = temp'*temp;
return
end
info(2) = temp'*temp;
return
end

2
matlab/CheckPath.m
View File

@ -2,7 +2,7 @@ function [DirectoryName, info] = CheckPath(type,dname)
% Creates the subfolder "./M_.dname/type" if it does not exist yet.
%
% INPUTS
% type [string] Name of the subfolder.
% type [string] Name of the subfolder.
% dname [string] Name of the directory
%
% OUTPUTS

2
matlab/CutSample.m
View File

@ -66,7 +66,7 @@ if (TotalNumberOfMhFiles-1)-(FirstMhFile+1)+1 > 0
elseif TotalNumberOfMhFiles == 1
record.KeepedDraws.Distribution = [];
elseif TotalNumberOfMhFiles == 2 && FirstMhFile > 1
record.KeepedDraws.Distribution = [MAX_nruns-FirstLine+1 ; record.MhDraws(end,3)];
record.KeepedDraws.Distribution = [MAX_nruns-FirstLine+1 ; record.MhDraws(end,3)];
end
% Save updated mh-history file.

84
matlab/DsgeSmoother.m
View File

@ -1,8 +1,8 @@
function [alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,T,R,P,PK,decomp,trend_addition,state_uncertainty,M_,oo_,options_,bayestopt_] = DsgeSmoother(xparam1,gend,Y,data_index,missing_value,M_,oo_,options_,bayestopt_,estim_params_)
% Estimation of the smoothed variables and innovations.
%
% INPUTS
% o xparam1 [double] (p*1) vector of (estimated) parameters.
% Estimation of the smoothed variables and innovations.
%
% INPUTS
% o xparam1 [double] (p*1) vector of (estimated) parameters.
% o gend [integer] scalar specifying the number of observations ==> varargin{1}.
% o data [double] (n*T) matrix of data.
% o data_index [cell] 1*smpl cell of column vectors of indices.
@ -12,7 +12,7 @@ function [alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,T,R,P,PK,de
% o options_ [structure] describing the options
% o bayestopt_ [structure] describing the priors
% o estim_params_ [structure] characterizing parameters to be estimated
%
%
% OUTPUTS
% o alphahat [double] (m*T) matrix, smoothed endogenous variables (a_{t|T}) (decision-rule order)
% o etahat [double] (r*T) matrix, smoothed structural shocks (r>=n is the number of shocks).
@ -29,31 +29,31 @@ function [alphahat,etahat,epsilonhat,ahat,SteadyState,trend_coeff,aK,T,R,P,PK,de
% matrices (meaningless for periods 1:d) (decision-rule order)
% o decomp (K*m*r*(T+K)) 4D array of shock decomposition of k-step ahead
% filtered variables (decision-rule order)
% o trend_addition [double] (n*T) pure trend component; stored in options_.varobs order
% o trend_addition [double] (n*T) pure trend component; stored in options_.varobs order
% o state_uncertainty [double] (K,K,T) array, storing the uncertainty
% about the smoothed state (decision-rule order)
% o M_ [structure] decribing the model
% o oo_ [structure] storing the results
% o options_ [structure] describing the options
% o bayestopt_ [structure] describing the priors
%
%
% Notes:
% m: number of endogenous variables (M_.endo_nbr)
% T: number of Time periods (options_.nobs)
% r: number of strucural shocks (M_.exo_nbr)
% n: number of observables (length(options_.varobs))
% K: maximum forecast horizon (max(options_.nk))
%
%
% To get variables that are stored in decision rule order in order of declaration
% as in M_.endo_names, ones needs code along the lines of:
% variables_declaration_order(dr.order_var,:) = alphahat
%
% Defines bayestopt_.mf = bayestopt_.smoother_mf (positions of observed variables
% and requested smoothed variables in decision rules (decision rule order)) and
%
% Defines bayestopt_.mf = bayestopt_.smoother_mf (positions of observed variables
% and requested smoothed variables in decision rules (decision rule order)) and
% passes it back via global variable
%
% ALGORITHM
% Diffuse Kalman filter (Durbin and Koopman)
%
% ALGORITHM
% Diffuse Kalman filter (Durbin and Koopman)
%
% SPECIAL REQUIREMENTS
% None
@ -97,7 +97,7 @@ end
%------------------------------------------------------------------------------
% 2. call model setup & reduction program
%------------------------------------------------------------------------------
oldoo.restrict_var_list = oo_.dr.restrict_var_list;
oldoo.restrict_var_list = oo_.dr.restrict_var_list;
oldoo.restrict_columns = oo_.dr.restrict_columns;
oo_.dr.restrict_var_list = bayestopt_.smoother_var_list;
oo_.dr.restrict_columns = bayestopt_.smoother_restrict_columns;
@ -133,8 +133,8 @@ mf = bayestopt_.mf;
% ------------------------------------------------------------------------------
% 3. Initial condition of the Kalman filter
% ------------------------------------------------------------------------------
%
% Here, Pinf and Pstar are determined. If the model is stationary, determine
%
% Here, Pinf and Pstar are determined. If the model is stationary, determine
% Pstar as the solution of the Lyapunov equation and set Pinf=[] (Notation follows
% Koopman/Durbin (2003), Journal of Time Series Analysis 24(1))
%
@ -169,7 +169,7 @@ elseif options_.lik_init == 3 % Diffuse Kalman filter
kalman_algo = 3;
else
if ~all(all(abs(H-diag(diag(H)))<1e-14))% ie, the covariance matrix is diagonal...
%Augment state vector (follows Section 6.4.3 of DK (2012))
%Augment state vector (follows Section 6.4.3 of DK (2012))
expanded_state_vector_for_univariate_filter=1;
T = blkdiag(T,zeros(vobs));
np = size(T,1);
@ -240,32 +240,32 @@ if kalman_algo == 1 || kalman_algo == 3
end
if kalman_algo == 2 || kalman_algo == 4
if ~all(all(abs(H-diag(diag(H)))<1e-14))% ie, the covariance matrix is diagonal...
if ~expanded_state_vector_for_univariate_filter
%Augment state vector (follows Section 6.4.3 of DK (2012))
expanded_state_vector_for_univariate_filter=1;
Z = [Z, eye(vobs)];
ST = blkdiag(ST,zeros(vobs));
np = size(ST,1);
Q = blkdiag(Q,H);
R1 = blkdiag(R,eye(vobs));
if kalman_algo == 4
%recompute Schur state space transformation with
%expanded state space
[Pstar,Pinf] = compute_Pinf_Pstar(mf,ST,R1,Q,options_.qz_criterium);
else
Pstar = blkdiag(Pstar,H);
if ~isempty(Pinf)
Pinf = blkdiag(Pinf,zeros(vobs));
end
end
%now reset H to 0
H = zeros(vobs,vobs);
if ~all(all(abs(H-diag(diag(H)))<1e-14))% ie, the covariance matrix is diagonal...
if ~expanded_state_vector_for_univariate_filter
%Augment state vector (follows Section 6.4.3 of DK (2012))
expanded_state_vector_for_univariate_filter=1;
Z = [Z, eye(vobs)];
ST = blkdiag(ST,zeros(vobs));
np = size(ST,1);
Q = blkdiag(Q,H);
R1 = blkdiag(R,eye(vobs));
if kalman_algo == 4
%recompute Schur state space transformation with
%expanded state space
[Pstar,Pinf] = compute_Pinf_Pstar(mf,ST,R1,Q,options_.qz_criterium);
else
%do nothing, state vector was already expanded
Pstar = blkdiag(Pstar,H);
if ~isempty(Pinf)
Pinf = blkdiag(Pinf,zeros(vobs));
end
end
%now reset H to 0
H = zeros(vobs,vobs);
else
%do nothing, state vector was already expanded
end
end
[alphahat,epsilonhat,etahat,ahat,P,aK,PK,decomp,state_uncertainty] = missing_DiffuseKalmanSmootherH3_Z(ST, ...
Z,R1,Q,diag(H), ...
Pinf,Pstar,data1,vobs,np,smpl,data_index, ...
@ -282,7 +282,7 @@ if expanded_state_vector_for_univariate_filter && (kalman_algo == 2 || kalman_al
ahat = ahat(k,:);
aK = aK(:,k,:,:);
epsilonhat=etahat(end-vobs+1:end,:);
etahat=etahat(1:end-vobs,:);
etahat=etahat(1:end-vobs,:);
if ~isempty(PK)
PK = PK(:,k,k,:);
end

8
matlab/GetAllPosteriorDraws.m
View File

@ -5,14 +5,14 @@ function Draws = GetAllPosteriorDraws(column, FirstMhFile, FirstLine, TotalNumbe
%
% INPUTS
% column: column
% FirstMhFile: first mh file
% FirstMhFile: first mh file
% FirstLine: first line
% TotalNumberOfMhFile: total number of mh file
% TotalNumberOfMhFile: total number of mh file
% NumberOfDraws: number of draws
% OUTPUTS
% Draws: draws from posterior distribution
%
%
% SPECIAL REQUIREMENTS
% none
@ -55,7 +55,7 @@ if nblck>1 && nargin<6
iline = 1;
end
end
else
else
for blck = 1:nblck
iline=iline0;
for file = FirstMhFile:TotalNumberOfMhFile

4
matlab/GetOneDraw.m
View File

@ -5,11 +5,11 @@ function [xparams, logpost] = GetOneDraw(type)
% INPUTS
% type: [string] 'posterior': draw from MCMC draws
% 'prior': draw from prior
%
%
% OUTPUTS
% xparams: vector of estimated parameters (drawn from posterior or prior distribution)
% logpost: log of the posterior density of this parameter vector
%
%
% SPECIAL REQUIREMENTS
% none

64
matlab/GetPosteriorMeanVariance.m
View File

@ -16,38 +16,38 @@ function [mean,variance] = GetPosteriorMeanVariance(M,drop)
%
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
MetropolisFolder = CheckPath('metropolis',M.dname);
FileName = M.fname;
BaseName = [MetropolisFolder filesep FileName];
load_last_mh_history_file(MetropolisFolder, FileName);
NbrDraws = sum(record.MhDraws(:,1));
NbrFiles = sum(record.MhDraws(:,2));
NbrBlocks = record.Nblck;
mean = 0;
variance = 0;
NbrKeptDraws = 0;
for i=1:NbrBlocks
NbrDrawsCurrentBlock = 0;
for j=1:NbrFiles
o = load([BaseName '_mh' int2str(j) '_blck' int2str(i),'.mat']);
NbrDrawsCurrentFile = size(o.x2,1);
if NbrDrawsCurrentBlock + NbrDrawsCurrentFile <= drop*NbrDraws
NbrDrawsCurrentBlock = NbrDrawsCurrentBlock + NbrDrawsCurrentFile;
continue
elseif NbrDrawsCurrentBlock < drop*NbrDraws
FirstDraw = ceil(drop*NbrDraws - NbrDrawsCurrentBlock + 1);
x2 = o.x2(FirstDraw:end,:);
else
x2 = o.x2;
end
NbrKeptDrawsCurrentFile = size(x2,1);
%recursively compute mean and variance
mean = (NbrKeptDraws*mean + sum(x2)')/(NbrKeptDraws+NbrKeptDrawsCurrentFile);
x2Demeaned = bsxfun(@minus,x2,mean');
variance = (NbrKeptDraws*variance + x2Demeaned'*x2Demeaned)/(NbrKeptDraws+NbrKeptDrawsCurrentFile);
MetropolisFolder = CheckPath('metropolis',M.dname);
FileName = M.fname;
BaseName = [MetropolisFolder filesep FileName];
load_last_mh_history_file(MetropolisFolder, FileName);
NbrDraws = sum(record.MhDraws(:,1));
NbrFiles = sum(record.MhDraws(:,2));
NbrBlocks = record.Nblck;
mean = 0;
variance = 0;
NbrKeptDraws = 0;
for i=1:NbrBlocks
NbrDrawsCurrentBlock = 0;
for j=1:NbrFiles
o = load([BaseName '_mh' int2str(j) '_blck' int2str(i),'.mat']);
NbrDrawsCurrentFile = size(o.x2,1);
if NbrDrawsCurrentBlock + NbrDrawsCurrentFile <= drop*NbrDraws
NbrDrawsCurrentBlock = NbrDrawsCurrentBlock + NbrDrawsCurrentFile;
NbrKeptDraws = NbrKeptDraws + NbrKeptDrawsCurrentFile;
continue
elseif NbrDrawsCurrentBlock < drop*NbrDraws
FirstDraw = ceil(drop*NbrDraws - NbrDrawsCurrentBlock + 1);
x2 = o.x2(FirstDraw:end,:);
else
x2 = o.x2;
end
NbrKeptDrawsCurrentFile = size(x2,1);
%recursively compute mean and variance
mean = (NbrKeptDraws*mean + sum(x2)')/(NbrKeptDraws+NbrKeptDrawsCurrentFile);
x2Demeaned = bsxfun(@minus,x2,mean');
variance = (NbrKeptDraws*variance + x2Demeaned'*x2Demeaned)/(NbrKeptDraws+NbrKeptDrawsCurrentFile);
NbrDrawsCurrentBlock = NbrDrawsCurrentBlock + NbrDrawsCurrentFile;
NbrKeptDraws = NbrKeptDraws + NbrKeptDrawsCurrentFile;
end
end

34
matlab/GetPosteriorParametersStatistics.m
View File

@ -1,17 +1,17 @@
function oo_ = GetPosteriorParametersStatistics(estim_params_, M_, options_, bayestopt_, oo_, pnames)
% This function prints and saves posterior estimates after the mcmc
% (+updates of oo_ & TeX output).
%
% INPUTS
% estim_params_ [structure]
% (+updates of oo_ & TeX output).
%
% INPUTS
% estim_params_ [structure]
% M_ [structure]
% options_ [structure]
% bayestopt_ [structure]
% oo_ [structure]
% pnames [char] Array of char, names of the prior shapes available
%
% OUTPUTS
% oo_ [structure]
%
% OUTPUTS
% oo_ [structure]
%
% SPECIAL REQUIREMENTS
% None.
@ -34,7 +34,7 @@ function oo_ = GetPosteriorParametersStatistics(estim_params_, M_, options_, bay
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
%if ~options_.mh_replic && options_.load_mh_file
% load([M_.fname '_results.mat'],'oo_');
% load([M_.fname '_results.mat'],'oo_');
%end
TeX = options_.TeX;
@ -48,7 +48,7 @@ nx = nvx+nvn+ncx+ncn+np;
MetropolisFolder = CheckPath('metropolis',M_.dname);
OutputFolder = CheckPath('Output',M_.dname);
FileName = M_.fname;
FileName = M_.fname;
load_last_mh_history_file(MetropolisFolder,FileName);
@ -108,14 +108,14 @@ if np
[post_mean, post_median, post_var, hpd_interval, post_deciles, ...
density] = posterior_moments(Draws,1,options_.mh_conf_sig);
name = bayestopt_.name{ip};
oo_ = Filloo(oo_,name,type,post_mean,hpd_interval,post_median,post_var,post_deciles,density);
oo_ = Filloo(oo_,name,type,post_mean,hpd_interval,post_median,post_var,post_deciles,density);
end
end
disp(sprintf(pformat,header_width,name,bayestopt_.p1(ip),...
post_mean, ...
hpd_interval, ...
pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip)));
bayestopt_.p2(ip)));
if TeX
k = estim_params_.param_vals(i,1);
name = deblank(M_.param_names_tex(k,:));
@ -171,7 +171,7 @@ if nvx
ip = ip+1;
end
if TeX
TeXEnd(fid,2,'standard deviation of structural shocks');
TeXEnd(fid,2,'standard deviation of structural shocks');
end
end
if nvn
@ -213,7 +213,7 @@ if nvn
ip = ip+1;
end
if TeX
TeXEnd(fid,3,'standard deviation of measurement errors');
TeXEnd(fid,3,'standard deviation of measurement errors');
end
end
if ncx
@ -314,12 +314,12 @@ if ncn
if TeX
name = ['(',deblank(M_.endo_names_tex(k1,:)) ',' deblank(M_.endo_names_tex(k2,:)),')'];
TeXCore(fid,name,deblank(pnames(bayestopt_.pshape(ip)+1,:)),bayestopt_.p1(ip),...
bayestopt_.p2(ip),post_mean,sqrt(post_var),hpd_interval);
bayestopt_.p2(ip),post_mean,sqrt(post_var),hpd_interval);
end
ip = ip+1;
end
if TeX
TeXEnd(fid,5,'correlation of measurement errors');
TeXEnd(fid,5,'correlation of measurement errors');
end
end
@ -359,7 +359,7 @@ fid = fidTeX;
function TeXCore(fid,name,shape,priormean,priorstd,postmean,poststd,hpd)
fprintf(fid,['$%s$ & %s & %7.3f & %6.4f & %7.3f& %6.4f & %7.4f & %7.4f \\\\ \n'],...
fprintf(fid,['$%s$ & %s & %7.3f & %6.4f & %7.3f& %6.4f & %7.4f & %7.4f \\\\ \n'],...
name,...
shape,...
priormean,...
@ -371,7 +371,7 @@ fprintf(fid,['$%s$ & %s & %7.3f & %6.4f & %7.3f& %6.4f & %7.4f & %7.4f \\\\ \n']
function TeXEnd(fid,fnum,title)
fprintf(fid,'\\end{longtable}\n ');
fprintf(fid,'\\end{longtable}\n ');
fprintf(fid,'\\end{center}\n');
fprintf(fid,'%% End of TeX file.\n');
fclose(fid);

12
matlab/MakeAllFigures.m
View File

@ -19,11 +19,11 @@ function MakeAllFigures(NumberOfPlots,Caption,FigureProperties,Info)
global M_ options_
FigHandle = figure('Name',FigureProperties.Name);
FigHandle = figure('Name',FigureProperties.Name);
NAMES = cell(NumberOfPlots,1);
if options_.TeX
TeXNAMES = cell(NumberOfPlots,1);
TeXNAMES = cell(NumberOfPlots,1);
end
if NumberOfPlots == 9
@ -53,7 +53,7 @@ elseif NumberOfPlots == 2
elseif NumberOfPlots == 1
nr = 1;
nc = 1;
end
end
for plt = 1:NumberOfPlots
eval(['NumberOfCurves = Info.Box' int2str(plt) '.Number;'])
@ -138,7 +138,7 @@ for plt = 1:NumberOfPlots
set(hh,'Color','r','LineStyle','-','LineWidth',2)
%
%
end
end
end
axis([xmin xmax ymin ymax])
title(NAMES{plt})
@ -150,14 +150,14 @@ if Info.SaveFormat.Eps
if isempty(Info.SaveFormat.Name)
eval(['print -depsc2 ' M_.fname Info.SaveFormat.GenericName int2str(Info.SaveFormat.Number) '.eps']);
else
eval(['print -depsc2 ' M_.fname Info.SaveFormat.GenericName Info.SaveFormat.Name '.eps']);
eval(['print -depsc2 ' M_.fname Info.SaveFormat.GenericName Info.SaveFormat.Name '.eps']);
end
end
if Info.SaveFormat.Pdf && ~isoctave
if isempty(Info.SaveFormat.Name)
eval(['print -dpdf ' M_.fname Info.SaveFormat.GenericName int2str(Info.SaveFormat.Number)]);
else
eval(['print -dpdf ' M_.fname Info.SaveFormat.GenericName Info.SaveFormat.Name]);
eval(['print -dpdf ' M_.fname Info.SaveFormat.GenericName Info.SaveFormat.Name]);
end
end
if Info.SaveFormat.Fig && ~isoctave

24
matlab/PlotPosteriorDistributions.m
View File

@ -4,15 +4,15 @@ function oo_ = PlotPosteriorDistributions(estim_params_, M_, options_, bayestopt
% plots posterior distributions
%
% INPUTS
% estim_params_ [structure]
% estim_params_ [structure]
% M_ [structure]
% options_ [structure]
% options_ [structure]
% bayestopt_ [structure]
% oo_ [structure]
%
%
% OUTPUTS
% oo_ [structure]
%
% oo_ [structure]
%
% SPECIAL REQUIREMENTS
% none
@ -49,7 +49,7 @@ nn = sqrt(MaxNumberOfPlotPerFigure);
figurename = 'Priors and posteriors';
if TeX && any(strcmp('eps',cellstr(options_.graph_format)))
if TeX && any(strcmp('eps',cellstr(options_.graph_format)))
fidTeX = fopen([OutputDirectoryName '/' M_.fname '_PriorsAndPosteriors.tex'],'w');
fprintf(fidTeX,'%% TeX eps-loader file generated by PlotPosteriorDistributions.m (Dynare).\n');
fprintf(fidTeX,['%% ' datestr(now,0) '\n']);
@ -78,9 +78,9 @@ for i=1:npar
end
end
[x2,f2,abscissa,dens,binf2,bsup2] = draw_prior_density(i,bayestopt_);
top2 = max(f2);
top2 = max(f2);
if i <= nvx
name = deblank(M_.exo_names(estim_params_.var_exo(i,1),:));
name = deblank(M_.exo_names(estim_params_.var_exo(i,1),:));
x1 = oo_.posterior_density.shocks_std.(name)(:,1);
f1 = oo_.posterior_density.shocks_std.(name)(:,2);
oo_.prior_density.shocks_std.(name)(:,1) = x2;
@ -96,18 +96,18 @@ for i=1:npar
oo_.prior_density.measurement_errors_std.(name)(:,2) = f2;
if ~options_.mh_posterior_mode_estimation
pmod = oo_.posterior_mode.measurement_errors_std.(name);
end
end
elseif i <= nvx+nvn+ncx
j = i - (nvx+nvn);
k1 = estim_params_.corrx(j,1);
k2 = estim_params_.corrx(j,2);
name = [deblank(M_.exo_names(k1,:)) '_' deblank(M_.exo_names(k2,:))];
name = [deblank(M_.exo_names(k1,:)) '_' deblank(M_.exo_names(k2,:))];
x1 = oo_.posterior_density.shocks_corr.(name)(:,1);
f1 = oo_.posterior_density.shocks_corr.(name)(:,2);
oo_.prior_density.shocks_corr.(name)(:,1) = x2;
oo_.prior_density.shocks_corr.(name)(:,2) = f2;
if ~options_.mh_posterior_mode_estimation
pmod = oo_.posterior_mode.shocks_corr.(name);
pmod = oo_.posterior_mode.shocks_corr.(name);
end
elseif i <= nvx+nvn+ncx+ncn
j = i - (nvx+nvn+ncx);
@ -157,7 +157,7 @@ for i=1:npar
fprintf(fidTeX,'\\begin{figure}[H]\n');
for j = 1:size(NAMES,1)
fprintf(fidTeX,'\\psfrag{%s}[1][][0.5][0]{%s}\n',deblank(NAMES(j,:)),deblank(TeXNAMES(j,:)));
end
end
fprintf(fidTeX,'\\centering\n');
fprintf(fidTeX,'\\includegraphics[width=%2.2f\\textwidth]{%s/%s_PriorsAndPosteriors%s}\n',options_.figures.textwidth*min(subplotnum/nn,1),OutputDirectoryName,M_.fname,int2str(figunumber));
fprintf(fidTeX,'\\caption{Priors and posteriors.}');

170
matlab/PosteriorIRF.m
View File

@ -368,101 +368,101 @@ end
% PosteriorIRF_core2.m function.
if ~options_.nograph && ~options_.no_graph.posterior
% Save the local variables.
localVars=[];
% Save the local variables.
localVars=[];
Check=options_.TeX;
if (Check)
localVars.varlist_TeX=varlist_TeX;
end
localVars.nvar=nvar;
localVars.MeanIRF=MeanIRF;
localVars.tit=tit;
localVars.nn=nn;
localVars.MAX_nirfs_dsgevar=MAX_nirfs_dsgevar;
localVars.HPDIRF=HPDIRF;
localVars.varlist=varlist;
localVars.MaxNumberOfPlotPerFigure=MaxNumberOfPlotPerFigure;
if options_.dsge_var
localVars.HPDIRFdsgevar=HPDIRFdsgevar;
localVars.MeanIRFdsgevar = MeanIRFdsgevar;
end
% The files .TeX are genereted in sequential way always!
% The files .TeX are generated in sequential way always!
subplotnum = 0;
tit_TeX(M_.exo_names_orig_ord,:) = M_.exo_names_tex;
if options_.TeX && any(strcmp('eps',cellstr(options_.graph_format)))
fidTeX = fopen([DirectoryName filesep M_.fname '_BayesianIRF.tex'],'w');
fprintf(fidTeX,'%% TeX eps-loader file generated by PosteriorIRF.m (Dynare).\n');
fprintf(fidTeX,['%% ' datestr(now,0) '\n']);
fprintf(fidTeX,' \n');
titTeX(M_.exo_names_orig_ord,:) = M_.exo_names_tex;
for ii=irf_shocks_indx
figunumber = 0;
for jj=1:nvar
if max(abs(MeanIRF(:,jj,ii))) >= options_.impulse_responses.plot_threshold
subplotnum = subplotnum+1;
if subplotnum == 1
fprintf(fidTeX,'\\begin{figure}[H]\n');
end
name = deblank(varlist(jj,:));
texname = deblank(varlist_TeX(jj,:));
fprintf(fidTeX,['\\psfrag{%s}[1][][0.5][0]{%s}\n'],name,['$' texname '$']);
end
if subplotnum == MaxNumberOfPlotPerFigure || (jj == nvar && subplotnum> 0)
figunumber = figunumber+1;
fprintf(fidTeX,'\\centering \n');
fprintf(fidTeX,'\\includegraphics[width=%2.2f\\textwidth]{%s/%s_Bayesian_IRF_%s_%d}\n',options_.figures.textwidth*min(subplotnum/nn,1),DirectoryName,M_.fname,deblank(tit(ii,:)),figunumber);
if options_.relative_irf
fprintf(fidTeX,['\\caption{Bayesian relative IRF.}']);
else
fprintf(fidTeX,'\\caption{Bayesian IRF: Orthogonalized shock to $%s$.}\n',deblank(tit_TeX(ii,:)));
end
fprintf(fidTeX,'\\label{Fig:BayesianIRF:%s:%d}\n',deblank(tit(ii,:)),figunumber);
fprintf(fidTeX,'\\end{figure}\n');
fprintf(fidTeX,' \n');
subplotnum = 0;
end
end
Check=options_.TeX;
if (Check)
localVars.varlist_TeX=varlist_TeX;
end
fprintf(fidTeX,'%% End of TeX file.\n');
fclose(fidTeX);
end
% The others file format are generated in parallel by PosteriorIRF_core2!
% Comment for testing!
if ~isoctave
if isnumeric(options_.parallel) || (M_.exo_nbr*ceil(size(varlist,1)/MaxNumberOfPlotPerFigure))<8
[fout] = PosteriorIRF_core2(localVars,1,M_.exo_nbr,0);
else
isRemoteOctave = 0;
for indPC=1:length(options_.parallel)
isRemoteOctave = isRemoteOctave + (findstr(options_.parallel(indPC).MatlabOctavePath, 'octave'));
localVars.nvar=nvar;
localVars.MeanIRF=MeanIRF;
localVars.tit=tit;
localVars.nn=nn;
localVars.MAX_nirfs_dsgevar=MAX_nirfs_dsgevar;
localVars.HPDIRF=HPDIRF;
localVars.varlist=varlist;
localVars.MaxNumberOfPlotPerFigure=MaxNumberOfPlotPerFigure;
if options_.dsge_var
localVars.HPDIRFdsgevar=HPDIRFdsgevar;
localVars.MeanIRFdsgevar = MeanIRFdsgevar;
end
% The files .TeX are genereted in sequential way always!
% The files .TeX are generated in sequential way always!
subplotnum = 0;
tit_TeX(M_.exo_names_orig_ord,:) = M_.exo_names_tex;
if options_.TeX && any(strcmp('eps',cellstr(options_.graph_format)))
fidTeX = fopen([DirectoryName filesep M_.fname '_BayesianIRF.tex'],'w');
fprintf(fidTeX,'%% TeX eps-loader file generated by PosteriorIRF.m (Dynare).\n');
fprintf(fidTeX,['%% ' datestr(now,0) '\n']);
fprintf(fidTeX,' \n');
titTeX(M_.exo_names_orig_ord,:) = M_.exo_names_tex;
for ii=irf_shocks_indx
figunumber = 0;
for jj=1:nvar
if max(abs(MeanIRF(:,jj,ii))) >= options_.impulse_responses.plot_threshold
subplotnum = subplotnum+1;
if subplotnum == 1
fprintf(fidTeX,'\\begin{figure}[H]\n');
end
name = deblank(varlist(jj,:));
texname = deblank(varlist_TeX(jj,:));
fprintf(fidTeX,['\\psfrag{%s}[1][][0.5][0]{%s}\n'],name,['$' texname '$']);
end
if subplotnum == MaxNumberOfPlotPerFigure || (jj == nvar && subplotnum> 0)
figunumber = figunumber+1;
fprintf(fidTeX,'\\centering \n');
fprintf(fidTeX,'\\includegraphics[width=%2.2f\\textwidth]{%s/%s_Bayesian_IRF_%s_%d}\n',options_.figures.textwidth*min(subplotnum/nn,1),DirectoryName,M_.fname,deblank(tit(ii,:)),figunumber);
if options_.relative_irf
fprintf(fidTeX,['\\caption{Bayesian relative IRF.}']);
else
fprintf(fidTeX,'\\caption{Bayesian IRF: Orthogonalized shock to $%s$.}\n',deblank(tit_TeX(ii,:)));
end
fprintf(fidTeX,'\\label{Fig:BayesianIRF:%s:%d}\n',deblank(tit(ii,:)),figunumber);
fprintf(fidTeX,'\\end{figure}\n');
fprintf(fidTeX,' \n');
subplotnum = 0;
end
end
end
if isRemoteOctave
fprintf(fidTeX,'%% End of TeX file.\n');
fclose(fidTeX);
end
% The others file format are generated in parallel by PosteriorIRF_core2!
% Comment for testing!
if ~isoctave
if isnumeric(options_.parallel) || (M_.exo_nbr*ceil(size(varlist,1)/MaxNumberOfPlotPerFigure))<8
[fout] = PosteriorIRF_core2(localVars,1,M_.exo_nbr,0);
else
globalVars = struct('M_',M_, ...
'options_', options_);
isRemoteOctave = 0;
for indPC=1:length(options_.parallel)
isRemoteOctave = isRemoteOctave + (findstr(options_.parallel(indPC).MatlabOctavePath, 'octave'));
end
if isRemoteOctave
[fout] = PosteriorIRF_core2(localVars,1,M_.exo_nbr,0);
else
globalVars = struct('M_',M_, ...
'options_', options_);
[fout] = masterParallel(options_.parallel, 1, M_.exo_nbr,NamFileInput,'PosteriorIRF_core2', localVars, globalVars, options_.parallel_info);
[fout] = masterParallel(options_.parallel, 1, M_.exo_nbr,NamFileInput,'PosteriorIRF_core2', localVars, globalVars, options_.parallel_info);
end
end
else
[fout] = PosteriorIRF_core2(localVars,1,M_.exo_nbr,0);
end
else
[fout] = PosteriorIRF_core2(localVars,1,M_.exo_nbr,0);
end
% END parallel code!
% END parallel code!
end

8
matlab/PosteriorIRF_core2.m
View File

@ -2,10 +2,10 @@ function myoutput=PosteriorIRF_core2(myinputs,fpar,npar,whoiam,ThisMatlab)
% function myoutput=PosteriorIRF_core2(myinputs,fpar,npar,whoiam, ThisMatlab)
% Generates the Posterior IRFs plot from the IRFs generated in
% PosteriorIRF_core1
%
%
% PARALLEL CONTEXT
% Performs in parallel execution a portion of the PosteriorIRF.m code.
% For more information, see the comment in posterior_sampler_core.m
% For more information, see the comment in posterior_sampler_core.m
% function.
%
% INPUTS
@ -135,7 +135,7 @@ for i=fpar:npar
plot(1:options_.irf,HPDIRFdsgevar(:,1,j,i),'--k','linewidth',1)
plot(1:options_.irf,HPDIRFdsgevar(:,2,j,i),'--k','linewidth',1)
end
% plot([1 options_.irf],[0 0],'-r','linewidth',0.5);
% plot([1 options_.irf],[0 0],'-r','linewidth',0.5);
box on
axis tight
xlim([1 options_.irf]);
@ -162,7 +162,7 @@ for i=fpar:npar
if whoiam
fprintf('Done! \n');
waitbarString = [ 'Exog. shocks ' int2str(i) '/' int2str(npar) ' done.'];
% fMessageStatus((i-fpar+1)/(npar-fpar+1),whoiam,waitbarString, waitbarTitle, Parallel(ThisMatlab));
% fMessageStatus((i-fpar+1)/(npar-fpar+1),whoiam,waitbarString, waitbarTitle, Parallel(ThisMatlab));
dyn_waitbar((i-fpar+1)/(npar-fpar+1),[],waitbarString);
end
end% loop over exo_var

14
matlab/ReshapeMatFiles.m
View File

@ -18,9 +18,9 @@ function ReshapeMatFiles(type, type2)
% posterior
% gsa
% prior
%
%
% OUTPUTS:
% none
% none
%
% SPECIAL REQUIREMENTS
% none
@ -61,17 +61,17 @@ else
end
else
MhDirectoryName = [CheckPath('prior',M_.dname) filesep ];
end
end
end
switch type
case 'irf_dsge'
CAPtype = 'IRF_DSGE';
TYPEsize = [ options_.irf , size(options_.varlist,1) , M_.exo_nbr ];
TYPEarray = 4;
TYPEarray = 4;
case 'irf_bvardsge'
CAPtype = 'IRF_BVARDSGE';
TYPEsize = [ options_.irf , length(options_.varobs) , M_.exo_nbr ];
TYPEarray = 4;
TYPEarray = 4;
case 'smooth'
CAPtype = 'SMOOTH';
TYPEsize = [ M_.endo_nbr , options_.nobs ];
@ -134,7 +134,7 @@ switch TYPEarray
eval(['idx = idx + size(stock_' type ',4);'])
end
%eval(['STOCK_' CAPtype ' = sort(STOCK_' CAPtype ',4);'])
save([MhDirectoryName M_.fname '_' CAPtype 's' int2str(NumberOfTYPEfiles-foffset+1) '.mat'],['STOCK_' CAPtype]);
save([MhDirectoryName M_.fname '_' CAPtype 's' int2str(NumberOfTYPEfiles-foffset+1) '.mat'],['STOCK_' CAPtype]);
end
else
load([MhDirectoryName M_.fname '_' type '1.mat']);
@ -176,7 +176,7 @@ switch TYPEarray
load([MhDirectoryName M_.fname '_' type '1.mat']);
%eval(['STOCK_' CAPtype ' = sort(stock_' type ',3);'])
eval(['STOCK_' CAPtype ' = stock_' type ';'])
save([MhDirectoryName M_.fname '_' CAPtype 's' int2str(1) '.mat'],['STOCK_' CAPtype ]);
save([MhDirectoryName M_.fname '_' CAPtype 's' int2str(1) '.mat'],['STOCK_' CAPtype ]);
end
% Original file format may be useful in some cases...
% for file = 1:NumberOfTYPEfiles

11
matlab/TaRB_optimizer_wrapper.m
View File

@ -3,15 +3,15 @@ function [fval,info,exit_flag,DLIK,Hess,SteadyState,trend_coeff] = TaRB_optimiz
% Wrapper function for target function used in TaRB algorithm; reassembles
% full parameter vector before calling target function
%
% INPUTS
% o optpar [double] (p_opt*1) vector of subset of parameters to be considered
% o par_vector [double] (p*1) full vector of parameters
% INPUTS
% o optpar [double] (p_opt*1) vector of subset of parameters to be considered
% o par_vector [double] (p*1) full vector of parameters
% o parameterindices [double] (p_opt*1) index of optpar entries in
% par_vector
% o TargetFun [char] string specifying the name of the objective
% function (posterior kernel).
% o varargin [structure] other inputs of target function
%
%
% OUTPUTS
% o fval [scalar] value of (minus) the likelihood.
% o info [double] (p*2) error code vector
@ -20,7 +20,7 @@ function [fval,info,exit_flag,DLIK,Hess,SteadyState,trend_coeff] = TaRB_optimiz
% o Hess [double] (p*p) asymptotic Hessian matrix.
% o SteadyState [double] Vector of doubles, steady state level for the endogenous variables.
% o trend_coeff [double] Matrix of doubles, coefficients of the deterministic trend in the measurement equation
%
%
% Copyright (C) 2015-16 Dynare Team
%
% This file is part of Dynare.
@ -40,4 +40,3 @@ function [fval,info,exit_flag,DLIK,Hess,SteadyState,trend_coeff] = TaRB_optimiz
par_vector(parameterindices,:)=optpar; %reassemble parameter
[fval,info,exit_flag,DLIK,Hess,SteadyState,trend_coeff] = feval(TargetFun,par_vector,varargin{:}); %call target function

6
matlab/Tracing.m
View File

@ -2,15 +2,15 @@ function [] = Tracing()
% DESCRIPTION
% This function is used to test the correct execution of a matlab section
% on remote machine.
%
%
% If no error happen the function simply create a file.
%
% INPUTS
% ...
%
%
% OUTPUTS
% ...
%
%
% SPECIAL REQUIREMENTS
% none

24
matlab/UnivariateSpectralDensity.m
View File

@ -1,15 +1,15 @@
function [oo_] = UnivariateSpectralDensity(M_,oo_,options_,var_list)
% This function computes the theoretical spectral density of each
% endogenous variable declared in var_list. Results are stored in
% oo_.SpectralDensity and may be plotted. Plots are saved into the
% graphs-folder.
%
% endogenous variable declared in var_list. Results are stored in
% oo_.SpectralDensity and may be plotted. Plots are saved into the
% graphs-folder.
%
% INPUTS
% M_ [structure] Dynare's model structure
% oo_ [structure] Dynare's results structure
% options_ [structure] Dynare's options structure
% var_list [integer] Vector of indices for a subset of variables.
%
%
% OUTPUTS
% oo_ [structure] Dynare's results structure,
% containing the subfield
@ -17,7 +17,7 @@ function [oo_] = UnivariateSpectralDensity(M_,oo_,options_,var_list)
% and density, which are of size nvar*ngrid.
%
% Adapted from th_autocovariances.m.
% Adapted from th_autocovariances.m.
% Copyright (C) 2006-2017 Dynare Team
%
@ -38,7 +38,7 @@ function [oo_] = UnivariateSpectralDensity(M_,oo_,options_,var_list)
if options_.order > 1
disp('UnivariateSpectralDensity :: I Cannot compute the theoretical spectral density')
disp('UnivariateSpectralDensity :: I Cannot compute the theoretical spectral density')
disp('with a second order approximation of the DSGE model!')
disp('Please set order = 1. I abort')
return
@ -102,7 +102,7 @@ if ~isempty(u)
ivar = oo_.dr.order_var(iky);
end
iky = iv(ivar);
iky = iv(ivar);
aa = ghx(iky,:);
bb = ghu(iky,:);
ngrid = options_.hp_ngrid; %number of grid points
@ -112,7 +112,7 @@ tneg = exp(-sqrt(-1)*freqs); %negative frequencies
if options_.one_sided_hp_filter
error('UnivariateSpectralDensity:: spectral density estimate not available with one-sided HP filter')
elseif options_.hp_filter == 0 && ~options_.bandpass.indicator %do not filter
filter_gain=ones(ngrid,1);
filter_gain=ones(ngrid,1);
elseif ~(options_.hp_filter == 0 && ~options_.bandpass.indicator) && options_.bandpass.indicator %filter with bandpass
filter_gain = zeros(1,ngrid);
lowest_periodicity=options_.bandpass.passband(2);
@ -122,7 +122,7 @@ elseif ~(options_.hp_filter == 0 && ~options_.bandpass.indicator) && options_.ba
filter_gain(freqs<=-2*pi/lowest_periodicity+2*pi & freqs>=-2*pi/highest_periodicity+2*pi)=1;
elseif ~(options_.hp_filter == 0 && ~options_.bandpass.indicator) && ~options_.bandpass.indicator %filter with HP-filter
lambda = options_.hp_filter;
filter_gain = 4*lambda*(1 - cos(freqs)).^2 ./ (1 + 4*lambda*(1 - cos(freqs)).^2);
filter_gain = 4*lambda*(1 - cos(freqs)).^2 ./ (1 + 4*lambda*(1 - cos(freqs)).^2);
end
mathp_col = NaN(ngrid,length(ivar)^2);
@ -139,7 +139,7 @@ end
f = zeros(nvar,ngrid);
for i=1:nvar
f(i,:) = real(mathp_col(:,(i-1)*nvar+i)); %read out spectral density
end
end
oo_.SpectralDensity.freqs=freqs;
oo_.SpectralDensity.density=f;
@ -164,7 +164,7 @@ if options_.nograph == 0
xlabel('0 \leq \omega \leq \pi')
ylabel('f(\omega)')
box on
axis tight
axis tight
dyn_saveas(hh,[M_.fname ,filesep,'graphs', filesep, 'SpectralDensity_' deblank(M_.endo_names(ivar(i),:))],options_.nodisplay,options_.graph_format)
end
end

25
matlab/WriteShockDecomp2Excel.m
View File

@ -1,7 +1,7 @@
function WriteShockDecomp2Excel(z,shock_names,endo_names,i_var,initial_date,DynareModel,DynareOptions,opts_decomp)
%function WriteShockDecomp2Excel(z,shock_names,endo_names,i_var,initial_date,DynareModel,DynareOptions)
% Saves the results from the shock_decomposition command to xls
%
%
% Inputs
% z [n_var*(nshock+2)*nperiods] shock decomposition array, see shock_decomposition.m for details
% shock_names [endo_nbr*string length] shock names from M_.exo_names
@ -48,8 +48,8 @@ if nargin==8
end
if isfield(opts_decomp,'fig_mode') && ~isempty(opts_decomp.fig_mode)
fig_mode = opts_decomp.fig_mode;
fig_mode1 = ['_' fig_mode];
fig_mode = [fig_mode '_'];
fig_mode1 = ['_' fig_mode];
fig_mode = [fig_mode '_'];
end
if isfield(opts_decomp,'screen_shocks')
if use_shock_groups
@ -60,15 +60,15 @@ if nargin==8
end
if isfield(opts_decomp,'fig_name')
fig_name = opts_decomp.fig_name;
% fig_name = ['_' fig_name];
fig_name1 = [fig_name];
fig_name = [fig_name '_'];
% fig_name = ['_' fig_name];
fig_name1 = [fig_name];
fig_name = [fig_name '_'];
end
if screen_shocks
fig_name1 = [fig_name1 '_screen'];
fig_name = [fig_name 'screen_'];
fig_name1 = [fig_name1 '_screen'];
fig_name = [fig_name 'screen_'];
end
end
end
gend = size(z,3);
@ -99,7 +99,7 @@ for j=1:nvar
z1 = [z1(isort(1:16),:); zres; z1(comp_nbr0:end,:)];
comp_nbr=18;
end
d0(1,:)=[{'Decomposition'} cellstr(labels(1:comp_nbr,:))' {'Smoot Var'}];
d0=[d0; num2cell([x' z1'])];
LastRow=size(d0,1);
@ -112,7 +112,7 @@ for j=1:nvar
d0(LastRow+2+ic,2:1+length(group_members))=group_members;
end
end
warning off
if ~ismac
[STATUS,MESSAGE] = xlswrite([DynareModel.fname,'_shock_decomposition',fig_mode,fig_name1],d0,deblank(endo_names(i_var(j),:)));
@ -122,6 +122,5 @@ for j=1:nvar
warning on
clear d0
end
end

4
matlab/add_filter_subtitle.m
View File

@ -1,7 +1,7 @@
function title=add_filter_subtitle(title,options_)
if ~options_.hp_filter && ~options_.one_sided_hp_filter && ~options_.bandpass.indicator %do not filter
%nothing to add here
%nothing to add here
elseif ~options_.hp_filter && ~options_.one_sided_hp_filter && options_.bandpass.indicator
title = [title ' (Bandpass filter, (' ...
num2str(options_.bandpass.passband(1)),' ',num2str(options_.bandpass.passband(2)), '))'];
@ -10,6 +10,6 @@ elseif options_.hp_filter && ~options_.one_sided_hp_filter && ~options_.bandpas
num2str(options_.hp_filter) ')'];
elseif ~options_.hp_filter && options_.one_sided_hp_filter && ~options_.bandpass.indicator
title = [title ' (One-sided HP filter, lambda = ' ...
num2str(options_.one_sided_hp_filter) ')'];
num2str(options_.one_sided_hp_filter) ')'];
end
end

4
matlab/add_path_to_mex_files.m
View File

@ -1,5 +1,5 @@
function mexpath = add_path_to_mex_files(dynareroot, modifypath)
% Copyright (C) 2015-2017 Dynare Team
%
% This file is part of Dynare.
@ -20,7 +20,7 @@ function mexpath = add_path_to_mex_files(dynareroot, modifypath)
if nargin<2
modifypath = true;
end
if exist('OCTAVE_VERSION')
if ispc() && strcmpi(computer(), 'i686-w64-mingw32')
mexpath = {[dynareroot '../mex/octave32/']};

237
matlab/annualized_shock_decomposition.m
View File

@ -1,28 +1,28 @@
function [z, endo_names, endo_names_tex, steady_state, i_var, oo_] = annualized_shock_decomposition(oo_, M_, options_, i_var, t0, t1, realtime_, vintage_, steady_state, q2a, cumfix)
% function oo_ = annualized_shock_decomposition(oo_,t0,options_.nobs);
% Computes annualized shocks contribution to a simulated trajectory. The fields set are
% oo_.annualized_shock_decomposition, oo_.annualized_realtime_shock_decomposition,
% oo_.annualized_realtime_conditional_shock_decomposition and oo_.annualized_realtime_forecast_shock_decomposition.
% oo_.annualized_shock_decomposition, oo_.annualized_realtime_shock_decomposition,
% oo_.annualized_realtime_conditional_shock_decomposition and oo_.annualized_realtime_forecast_shock_decomposition.
% Subfields are arrays n_var by nshock+2 by nperiods. The
% first nshock columns store the respective shock contributions, column n+1
% stores the role of the initial conditions, while column n+2 stores the
% value of the smoothed variables. Both the variables and shocks are stored
% value of the smoothed variables. Both the variables and shocks are stored
% in the order of endo_names and M_.exo_names, respectively.
%
% INPUTS
% oo_: [structure] Storage of results
% M_: [structure] Storage of model
% opts: [structure] options for shock decomp
% opts: [structure] options for shock decomp
% i_var: [array] index of vars
% t0: [integer] first period
% t1: [integer] last period
% realtime_: [integer]
% realtime_: [integer]
% vintage_: [integer]
% steady_state: [array] steady state value of quarterly (log-) level vars
% q2a: [structure] info on q2a
%
% OUTPUTS
% z: [matrix] shock decomp to plot
% z: [matrix] shock decomp to plot
% endo_names: [char] updated var names
% endo_names_tex: [char] updated TeX var names
% steady_state: [array] updated stady state of vars
@ -57,13 +57,13 @@ islog = q2a.islog;
aux = q2a.aux;
aux0 = aux;
cumfix = q2a.cumfix;
% usual shock decomp
% usual shock decomp
if isstruct(oo_)
% z = oo_.shock_decomposition;
myopts=options_;
myopts.plot_shock_decomp.type='qoq';
myopts.plot_shock_decomp.realtime=0;
[z, junk] = plot_shock_decomposition(M_,oo_,myopts,[]);
% z = oo_.shock_decomposition;
myopts=options_;
myopts.plot_shock_decomp.type='qoq';
myopts.plot_shock_decomp.realtime=0;
[z, junk] = plot_shock_decomposition(M_,oo_,myopts,[]);
else
z = oo_;
end
@ -177,147 +177,147 @@ end
% realtime
if realtime_ && isstruct(oo_) && isfield(oo_, 'realtime_shock_decomposition')
init=1;
for i=t0:4:t1
yr=floor(i/4);
za=[];
gza=[];
init=1;
for i=t0:4:t1
yr=floor(i/4);
za=[];
gza=[];
myopts=options_;
myopts.plot_shock_decomp.type='qoq';
myopts.plot_shock_decomp.realtime=1;
myopts.plot_shock_decomp.vintage=i;
[z, steady_state_aux] = plot_shock_decomposition(M_,oo_,myopts,[]);
z = z(i_var,:,:);
if isstruct(aux)
if ischar(aux0.y)
[y_aux, steady_state_aux] = plot_shock_decomposition(M_,oo_,myopts,aux0.y);
aux.y=y_aux;
aux.yss=steady_state_aux;
end
yaux=aux.y;
end
nterms = size(z,2);
% z = oo_.realtime_shock_decomposition.(['time_' int2str(i)]);
% z = z(i_var,:,:);
for j=1:nvar
for k =nterms:-1:1
% if k<nterms
% ztmp = squeeze(sum(z(j,[1:k-1,k+1:end-1],t0-4:end)));
% else
ztmp = squeeze(z(j,k,min((t0-3):-4:1):end));
% end
if isstruct(aux)
aux.y = squeeze(yaux(j,k,min((t0-3):-4:1):end));
if isstruct(aux)
if ischar(aux0.y)
[y_aux, steady_state_aux] = plot_shock_decomposition(M_,oo_,myopts,aux0.y);
aux.y=y_aux;
aux.yss=steady_state_aux;
end
[za(j,k,:), steady_state_a(j,1), gza(j,k,:), steady_state_ga(j,1)] = ...
quarterly2annual(ztmp,steady_state(j),GYTREND0,var_type,islog,aux);
% if k<nterms
% za(j,k,:) = za(j,end,:) - za(j,k,:);
% gza(j,k,:) = gza(j,end,:) - gza(j,k,:);
% end
yaux=aux.y;
end
ztmp=squeeze(za(j,:,:));
nterms = size(z,2);
if cumfix==0
zscale = sum(ztmp(1:end-1,:))./ztmp(end,:);
ztmp(1:end-1,:) = ztmp(1:end-1,:)./repmat(zscale,[nterms-1,1]);
else
zres = ztmp(end,:)-sum(ztmp(1:end-1,:));
ztmp(end-1,:) = ztmp(end-1,:) + zres;
% z = oo_.realtime_shock_decomposition.(['time_' int2str(i)]);
% z = z(i_var,:,:);
for j=1:nvar
for k =nterms:-1:1
% if k<nterms
% ztmp = squeeze(sum(z(j,[1:k-1,k+1:end-1],t0-4:end)));
% else
ztmp = squeeze(z(j,k,min((t0-3):-4:1):end));
% end
if isstruct(aux)
aux.y = squeeze(yaux(j,k,min((t0-3):-4:1):end));
end
[za(j,k,:), steady_state_a(j,1), gza(j,k,:), steady_state_ga(j,1)] = ...
quarterly2annual(ztmp,steady_state(j),GYTREND0,var_type,islog,aux);
% if k<nterms
% za(j,k,:) = za(j,end,:) - za(j,k,:);
% gza(j,k,:) = gza(j,end,:) - gza(j,k,:);
% end
end
ztmp=squeeze(za(j,:,:));
if cumfix==0
zscale = sum(ztmp(1:end-1,:))./ztmp(end,:);
ztmp(1:end-1,:) = ztmp(1:end-1,:)./repmat(zscale,[nterms-1,1]);
else
zres = ztmp(end,:)-sum(ztmp(1:end-1,:));
ztmp(end-1,:) = ztmp(end-1,:) + zres;
end
gztmp=squeeze(gza(j,:,:));
if cumfix==0
gscale = sum(gztmp(1:end-1,:))./ gztmp(end,:);
gztmp(1:end-1,:) = gztmp(1:end-1,:)./repmat(gscale,[nterms-1,1]);
else
gres = gztmp(end,:) - sum(gztmp(1:end-1,:));
gztmp(end-1,:) = gztmp(end-1,:)+gres;
end
za(j,:,:) = ztmp;
gza(j,:,:) = gztmp;
end
gztmp=squeeze(gza(j,:,:));
if cumfix==0
gscale = sum(gztmp(1:end-1,:))./ gztmp(end,:);
gztmp(1:end-1,:) = gztmp(1:end-1,:)./repmat(gscale,[nterms-1,1]);
if q2a.plot ==1
z=gza;
elseif q2a.plot == 2
z=za;
else
gres = gztmp(end,:) - sum(gztmp(1:end-1,:));
gztmp(end-1,:) = gztmp(end-1,:)+gres;
z=cat(1,za,gza);
end
za(j,:,:) = ztmp;
gza(j,:,:) = gztmp;
end
if q2a.plot ==1
z=gza;
elseif q2a.plot == 2
z=za;
else
z=cat(1,za,gza);
end
if init==1
oo_.annualized_realtime_shock_decomposition.pool = z;
else
oo_.annualized_realtime_shock_decomposition.pool(:,:,yr) = z(:,:,end-nfrcst);
end
oo_.annualized_realtime_shock_decomposition.(['yr_' int2str(yr)]) = z;
if opts.forecast
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr)]) = z(:,:,end-nfrcst:end);
if init>nfrcst
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)]) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,:,yr-nfrcst:end) - ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-nfrcst)]);
% fix others
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end-1,:) = ...
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end-1,:) + ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end,:);
% fix total
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end,:) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,end,yr-nfrcst:end);
if i==t1
for my_forecast_=(nfrcst-1):-1:1
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-my_forecast_)]) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,:,yr-my_forecast_:yr) - ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,:,1:my_forecast_+1);
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,end-1,:) = ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,end,1:my_forecast_+1);
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,end,:) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,end,yr-my_forecast_:yr);
if init==1
oo_.annualized_realtime_shock_decomposition.pool = z;
else
oo_.annualized_realtime_shock_decomposition.pool(:,:,yr) = z(:,:,end-nfrcst);
end
oo_.annualized_realtime_shock_decomposition.(['yr_' int2str(yr)]) = z;
if opts.forecast
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr)]) = z(:,:,end-nfrcst:end);
if init>nfrcst
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)]) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,:,yr-nfrcst:end) - ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-nfrcst)]);
% fix others
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end-1,:) = ...
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end-1,:) + ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end,:);
% fix total
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-nfrcst)])(:,end,:) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,end,yr-nfrcst:end);
if i==t1
for my_forecast_=(nfrcst-1):-1:1
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-my_forecast_)]) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,:,yr-my_forecast_:yr) - ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,:,1:my_forecast_+1);
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,end-1,:) = ...
oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,end,1:my_forecast_+1);
oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(yr-my_forecast_)])(:,end,:) = ...
oo_.annualized_realtime_shock_decomposition.pool(:,end,yr-my_forecast_:yr);
end
end
end
end
% ztmp=oo_.realtime_shock_decomposition.pool(:,:,21:29)-oo_.realtime_forecast_shock_decomposition.time_21;
init=init+1;
end
% ztmp=oo_.realtime_shock_decomposition.pool(:,:,21:29)-oo_.realtime_forecast_shock_decomposition.time_21;
switch realtime_
init=init+1;
end
switch realtime_
case 0
case 0
z = oo_.annualized_shock_decomposition;
case 1 % realtime
case 1 % realtime
if vintage_
z = oo_.annualized_realtime_shock_decomposition.(['yr_' int2str(floor(vintage_/4))]);
else
z = oo_.annualized_realtime_shock_decomposition.pool;
end
case 2 % conditional
case 2 % conditional
if vintage_
z = oo_.annualized_realtime_conditional_shock_decomposition.(['yr_' int2str(floor(vintage_/4))]);
else
error();
end
case 3 % forecast
case 3 % forecast
if vintage_
z = oo_.annualized_realtime_forecast_shock_decomposition.(['yr_' int2str(floor(vintage_/4))]);
else
error()
end
end
end
end
if q2a.plot ==0
@ -331,4 +331,3 @@ else
steady_state = steady_state_a;
end
end

24
matlab/autoregressive_process_specification.m
View File

@ -2,43 +2,43 @@ function [InnovationVariance,AutoregressiveParameters] = autoregressive_process_
% This function computes the parameters of an AR(p) process from the variance and the autocorrelation function
% (the first p terms) of this process.
%
% INPUTS
% INPUTS
% [1] Variance [double] scalar, variance of the variable.
% [2] Rho [double] p*1 vector, the autocorelation function: \rho(1), \rho(2), ..., \rho(p).
% [3] p [double] scalar, the number of lags in the AR process.
%
% OUTPUTS
% OUTPUTS
% [1] InnovationVariance [double] scalar, the variance of the innovation.
% [2] AutoregressiveParameters [double] p*1 vector of autoregressive parameters.
%
% NOTES
% NOTES
%
% The AR(p) model for {y_t} is:
%
% y_t = \phi_1 * y_{t-1} + \phi_2 * y_{t-2} + ... + \phi_p * y_{t-p} + e_t
%
% y_t = \phi_1 * y_{t-1} + \phi_2 * y_{t-2} + ... + \phi_p * y_{t-p} + e_t
%
% Let \gamma(0) and \rho(1), ..., \rho(2) be the variance and the autocorrelation function of {y_t}. This function
% compute the variance of {e_t} and the \phi_i (i=1,...,p) from the variance and the autocorrelation function of {y_t}.
% compute the variance of {e_t} and the \phi_i (i=1,...,p) from the variance and the autocorrelation function of {y_t}.
% We know that:
%
%
% \gamma(0) = \phi_1 \gamma(1) + ... + \phi_p \gamma(p) + \sigma^2
%
% where \sigma^2 is the variance of {e_t}. Equivalently we have:
%
% \sigma^2 = \gamma(0) (1-\rho(1)\phi_1 - ... - \rho(p)\phi_p)
% \sigma^2 = \gamma(0) (1-\rho(1)\phi_1 - ... - \rho(p)\phi_p)
%
% We also have for any integer h>0:
%
%
% \rho(h) = \phi_1 \rho(h-1) + ... + \phi_p \rho(h-p)
%
% We can write the equations for \rho(1), ..., \rho(p) using matrices. Let R be the p*p autocorelation
% matrix and v be the p*1 vector gathering the first p terms of the autocorrelation function. We have:
% matrix and v be the p*1 vector gathering the first p terms of the autocorrelation function. We have:
%
% v = R*PHI
%
%
% where PHI is a p*1 vector with the autoregressive parameters of the AR(p) process. We can recover the autoregressive
% parameters by inverting the autocorrelation matrix: PHI = inv(R)*v.
%
%
% This function first computes the vector PHI by inverting R and computes the variance of the innovation by evaluating
%
% \sigma^2 = \gamma(0)*(1-PHI'*v)

6
matlab/backward_model_inversion.m
View File

@ -1,6 +1,6 @@
function [endogenousvariables, exogenousvariables] = backward_model_inversion(constraints, exogenousvariables, initialconditions, endo_names, exo_names, freeinnovations, DynareModel, DynareOptions, DynareOutput)
% INPUTS
% INPUTS
% - constraints [dseries] with N constrained endogenous variables from t1 to t2.
% - exogenousvariables [dseries] with Q exogenous variables.
% - initialconditions [dseries] with M endogenous variables starting before t1 (M initialcond must contain at least the state variables).
@ -8,11 +8,11 @@ function [endogenousvariables, exogenousvariables] = backward_model_inversion(co
% - exo_names [cell] list of exogenous variable names.
% - freeinstruments [cell] list of exogenous variable names used to control the constrained endogenous variables.
%
% OUTPUTS
% OUTPUTS
% - endogenous [dseries]
% - exogenous [dseries]
%
% REMARKS
% REMARKS
% Copyright (C) 2017 Dynare Team
%

114
matlab/basic_plan.m
View File

@ -31,60 +31,60 @@ function plan = basic_plan(plan, exogenous, expectation_type, date, value)
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
if ~ischar(expectation_type) || size(expectation_type,1) ~= 1
error(['in basic_plan the third argument should be a string containing the simulation type (''perfect_foresight'' or ''surprise'')']);
end
exogenous = strtrim(exogenous);
ix = find(strcmp(exogenous, plan.exo_names));
if isempty(ix)
error(['in basic_plan the second argument ' exogenous ' is not an exogenous variable']);
end
sdate = length(date);
if sdate > 1
if date(1) < plan.date(1) || date(end) > plan.date(end)
error(['in basic_plan the fourth argument (date=' date ') must lay inside the plan.date ' plan.date]);
end
else
if date < plan.date(1) || date > plan.date(end)
error(['in basic_plan the fourth argument (date=' date ') must lay iside the plan.date ' plan.date]);
end
end
if length(date) ~= length(value)
error(['in basic_plan the number of dates (' int2str(length(date)) ') is not equal to the numbers of shock (' int2str(length(value)) ') for exogenous variable ' exogenous]);
end
if ~isempty(plan.options_cond_fcst_.controlled_varexo)
common_var = find(ix == plan.options_cond_fcst_.controlled_varexo);
if ~isempty(common_var)
common_date = intersect(date, plan.constrained_date_{common_var});
if ~isempty(common_date)
[date_, i_date] = setdiff(date, common_date);
value = value(i_date);
if common_date.length > 1
the_dates = [cell2mat(strings(common_date(1))) ':' cell2mat(strings(common_date(end)))];
else
the_dates = cell2mat(strings(common_date));
end
warning(['Impossible case: ' plan.exo_names{plan.options_cond_fcst_.controlled_varexo(common_var)} ' is used both as a shock and as an endogenous variable to control the path of ' plan.endo_names{plan.constrained_vars_(common_var)} ' at the dates ' the_dates]);
warning('This shock will not be considered');
end
end
end
if isempty(plan.shock_vars_)
plan.shock_vars_ = ix;
if strcmp(expectation_type, 'perfect_foresight')
plan.shock_perfect_foresight_ = 1;
else
plan.shock_perfect_foresight_ = 0;
end
else
plan.shock_vars_ = [plan.shock_vars_ ; ix];
if strcmp(expectation_type, 'perfect_foresight')
plan.shock_perfect_foresight_ = [plan.shock_perfect_foresight_ ; 1];
else
plan.shock_perfect_foresight_ = [plan.shock_perfect_foresight_ ; 0];
end
end
plan.shock_date_{length(plan.shock_date_) + 1} = date;
plan.shock_str_date_{length(plan.shock_str_date_) + 1} = strings(date);
plan.shock_int_date_{length(plan.shock_int_date_) + 1} = date - plan.date(1) + 1;
plan.shock_paths_{length(plan.shock_paths_) + 1} = value;
if ~ischar(expectation_type) || size(expectation_type,1) ~= 1
error(['in basic_plan the third argument should be a string containing the simulation type (''perfect_foresight'' or ''surprise'')']);
end
exogenous = strtrim(exogenous);
ix = find(strcmp(exogenous, plan.exo_names));
if isempty(ix)
error(['in basic_plan the second argument ' exogenous ' is not an exogenous variable']);
end
sdate = length(date);
if sdate > 1
if date(1) < plan.date(1) || date(end) > plan.date(end)
error(['in basic_plan the fourth argument (date=' date ') must lay inside the plan.date ' plan.date]);
end
else
if date < plan.date(1) || date > plan.date(end)
error(['in basic_plan the fourth argument (date=' date ') must lay iside the plan.date ' plan.date]);
end
end
if length(date) ~= length(value)
error(['in basic_plan the number of dates (' int2str(length(date)) ') is not equal to the numbers of shock (' int2str(length(value)) ') for exogenous variable ' exogenous]);
end
if ~isempty(plan.options_cond_fcst_.controlled_varexo)
common_var = find(ix == plan.options_cond_fcst_.controlled_varexo);
if ~isempty(common_var)
common_date = intersect(date, plan.constrained_date_{common_var});
if ~isempty(common_date)
[date_, i_date] = setdiff(date, common_date);
value = value(i_date);
if common_date.length > 1
the_dates = [cell2mat(strings(common_date(1))) ':' cell2mat(strings(common_date(end)))];
else
the_dates = cell2mat(strings(common_date));
end
warning(['Impossible case: ' plan.exo_names{plan.options_cond_fcst_.controlled_varexo(common_var)} ' is used both as a shock and as an endogenous variable to control the path of ' plan.endo_names{plan.constrained_vars_(common_var)} ' at the dates ' the_dates]);
warning('This shock will not be considered');
end
end
end
if isempty(plan.shock_vars_)
plan.shock_vars_ = ix;
if strcmp(expectation_type, 'perfect_foresight')
plan.shock_perfect_foresight_ = 1;
else
plan.shock_perfect_foresight_ = 0;
end
else
plan.shock_vars_ = [plan.shock_vars_ ; ix];
if strcmp(expectation_type, 'perfect_foresight')
plan.shock_perfect_foresight_ = [plan.shock_perfect_foresight_ ; 1];
else
plan.shock_perfect_foresight_ = [plan.shock_perfect_foresight_ ; 0];
end
end
plan.shock_date_{length(plan.shock_date_) + 1} = date;
plan.shock_str_date_{length(plan.shock_str_date_) + 1} = strings(date);
plan.shock_int_date_{length(plan.shock_int_date_) + 1} = date - plan.date(1) + 1;
plan.shock_paths_{length(plan.shock_paths_) + 1} = value;

2
matlab/bksup0.m
View File

@ -4,7 +4,7 @@ function d = bksup0(c,ny,jcf,iyf,icf,periods)
% INPUTS
% ny: number of endogenous variables
% jcf: variables index forward
%
%
% OUTPUTS
% d: vector of backsubstitution results
%

3
matlab/bksup1.m
View File

@ -5,7 +5,7 @@ function d = bksup1(c,ny,jcf,iyf,periods)
% INPUTS
% ny: number of endogenous variables
% jcf: variables index forward
%
%
% OUTPUTS
% d: vector of backsubstitution results
%
@ -40,4 +40,3 @@ for i = 2:periods
end
d = c(:,jcf) ;

2
matlab/bksupk.m
View File

@ -68,7 +68,7 @@ while i <= options_.periods
c = fread(fid,[jcf,ny],'float64')' ;
d1(ir) = c(:,jcf)-c(:,icf)*d1(irf) ;
ir = ir-ny ;
ir = ir-ny ;
irf = irf-ny ;
i = i+1;
end

7
matlab/bseastr.m
View File

@ -33,16 +33,15 @@ for im = 1:m
for i = 1:min(length(key),length(temp))
if temp(i) > key(i)
h = mid - 1 ;
break
break
else
l = mid + 1 ;
break
break
end
end
else
x(im) = mid ;
break
break
end
end
end

12
matlab/bvar_density.m
View File

@ -37,16 +37,16 @@ for nlags = 1:maxnlags
[ny, nx, posterior, prior] = bvar_toolbox(nlags);
oo_.bvar.posterior{nlags}=posterior;
oo_.bvar.prior{nlags}=prior;
posterior_int = matrictint(posterior.S, posterior.df, posterior.XXi);
prior_int = matrictint(prior.S, prior.df, prior.XXi);
lik_nobs = posterior.df - prior.df;
log_dnsty = posterior_int - prior_int - 0.5*ny*lik_nobs*log(2*pi);
oo_.bvar.log_marginal_data_density(nlags)=log_dnsty;
skipline()
fprintf('The marginal log density of the BVAR(%g) model is equal to %10.4f\n', ...
nlags, log_dnsty);
@ -61,7 +61,7 @@ function w = matrictint(S, df, XXi)
% S: parameter of inverse-Wishart distribution
% df: number of degrees of freedom of inverse-Wishart distribution
% XXi: first component of VCV matrix of matrix-normal distribution
%
%
% Computes the integral over (Phi, Sigma) of:
%
% det(Sigma)^(-k/2)*exp(-0.5*Tr((Phi-PhiHat)'*(XXi)^(-1)*(Phi-PhiHat)*Sigma^(-1)))*

28
matlab/bvar_forecast.m
View File

@ -54,7 +54,7 @@ p = 0;
% Loop counter initialization
d = 0;
while d <= options_.bvar_replic
Sigma = rand_inverse_wishart(ny, posterior.df, S_inv_upper_chol);
% Option 'lower' of chol() not available in old versions of
@ -62,15 +62,15 @@ while d <= options_.bvar_replic
Sigma_lower_chol = chol(Sigma)';
Phi = rand_matrix_normal(k, ny, posterior.PhiHat, Sigma_lower_chol, XXi_lower_chol);
% All the eigenvalues of the companion matrix have to be on or inside the unit circle
Companion_matrix(1:ny,:) = Phi(1:ny*nlags,:)';
Companion_matrix(1:ny,:) = Phi(1:ny*nlags,:)';
test = (abs(eig(Companion_matrix)));
if any(test>1.0000000000001)
p = p+1;
end
d = d+1;
% Without shocks
lags_data = forecast_data.initval;
for t = 1:options_.forecast
@ -80,7 +80,7 @@ while d <= options_.bvar_replic
lags_data(end,:) = y;
sims_no_shock(t, :, d) = y;
end
% With shocks
lags_data = forecast_data.initval;
for t = 1:options_.forecast
@ -120,9 +120,9 @@ dyn_graph=dynare_graph_init(sprintf('BVAR forecasts (nlags = %d)', nlags), ny, {
for i = 1:ny
dyn_graph=dynare_graph(dyn_graph,[ sims_no_shock_median(:, i) ...
sims_no_shock_up_conf(:, i) sims_no_shock_down_conf(:, i) ...
sims_with_shocks_up_conf(:, i) sims_with_shocks_down_conf(:, i) ], ...
options_.varobs{i});
sims_no_shock_up_conf(:, i) sims_no_shock_down_conf(:, i) ...
sims_with_shocks_up_conf(:, i) sims_with_shocks_down_conf(:, i) ], ...
options_.varobs{i});
end
dyn_saveas(dyn_graph.fh,[OutputDirectoryName '/' M_.fname '_BVAR_forecast_',num2str(nlags)],options_.nodisplay,options_.graph_format)
@ -130,9 +130,9 @@ dyn_saveas(dyn_graph.fh,[OutputDirectoryName '/' M_.fname '_BVAR_forecast_',num2
% Compute RMSE
if ~isempty(forecast_data.realized_val)
sq_err_cumul = zeros(1, ny);
lags_data = forecast_data.initval;
for t = 1:size(forecast_data.realized_val, 1)
X = [ reshape(flipdim(lags_data, 1)', 1, ny*nlags) forecast_data.realized_xdata(t, :) ];
@ -141,14 +141,14 @@ if ~isempty(forecast_data.realized_val)
lags_data(end,:) = y;
sq_err_cumul = sq_err_cumul + (y - forecast_data.realized_val(t, :)) .^ 2;
end
rmse = sqrt(sq_err_cumul / size(forecast_data.realized_val, 1));
fprintf('RMSE of BVAR(%d):\n', nlags);
for i = 1:length(options_.varobs)
fprintf('%s: %10.4f\n', options_.varobs{i}, rmse(i));
end
end
end
% Store results

14
matlab/bvar_irf.m
View File

@ -54,7 +54,7 @@ p = 0;
sampled_irfs = NaN(ny, ny, options_.irf, options_.bvar_replic);
for draw=1:options_.bvar_replic
% Get a covariance matrix from an inverted Wishart distribution.
Sigma = rand_inverse_wishart(ny, posterior.df, S_inv_upper_chol);
Sigma_upper_chol = chol(Sigma);
@ -62,10 +62,10 @@ for draw=1:options_.bvar_replic
% Get the Autoregressive matrices from a matrix variate distribution.
Phi = rand_matrix_normal(k, ny, posterior.PhiHat, Sigma_lower_chol, XXi_lower_chol);
% Form the companion matrix.
Companion_matrix(1:ny,:) = transpose(Phi(1:ny*nlags,:));
Companion_matrix(1:ny,:) = transpose(Phi(1:ny*nlags,:));
% All the eigenvalues of the companion matrix have to be on or
% inside the unit circle to rule out explosive time series.
test = (abs(eig(Companion_matrix)));
@ -78,7 +78,7 @@ for draw=1:options_.bvar_replic
elseif strcmpi(identification,'SquareRoot')
StructuralMat = sqrtm(Sigma);
end
% Build the IRFs...
lags_data = zeros(ny,ny*nlags) ;
sampled_irfs(:,:,1,draw) = Sigma_lower_chol ;
@ -88,7 +88,7 @@ for draw=1:options_.bvar_replic
lags_data(:,ny+1:end) = lags_data(:,1:end-ny) ;
lags_data(:,1:ny) = sampled_irfs(:,:,t,draw) ;
end
end
if p > 0
@ -106,7 +106,7 @@ sort_idx = round((0.5 + [-options_.bvar.conf_sig, options_.bvar.conf_sig, .0]/2)
posterior_down_conf_irfs = sorted_irfs(:,:,:,sort_idx(1));
posterior_up_conf_irfs = sorted_irfs(:,:,:,sort_idx(2));
posterior_median_irfs = sorted_irfs(:,:,:,sort_idx(3));
posterior_median_irfs = sorted_irfs(:,:,:,sort_idx(3));
number_of_columns = fix(sqrt(ny));
number_of_rows = ceil(ny / number_of_columns) ;

18
matlab/bvar_toolbox.m
View File

@ -29,7 +29,7 @@ function [ny, nx, posterior, prior, forecast_data] = bvar_toolbox(nlags)
% forecasting, of size options_.forecast*nx (actually only
% contains "1" values for the constant term if nx ~= 0)
% - realized_val: only non-empty if options_.nobs doesn't point
% to the end of sample
% to the end of sample
% In that case, contains values of endogenous variables after
% options_.nobs and up to the end of the sample
% - realized_xdata: contains values of exogenous variables after
@ -67,7 +67,7 @@ options_ = set_default_option(options_, 'nobs', size(dataset,1)-options_.first_o
if (options_.first_obs+options_.nobs-1)> size(dataset,1)
fprintf('Incorrect or missing specification of the number of observations. nobs can be at most %4u\n',size(dataset,1)-options_.first_obs+1);
error('Inconsistent number of observations.')
error('Inconsistent number of observations.')
end
% Parameters for prior
@ -160,7 +160,7 @@ function [ydum,xdum,breaks]=varprior(nv,nx,lags,mnprior,vprior)
%function [ydum,xdum,breaks]=varprior(nv,nx,lags,mnprior,vprior)
% ydum, xdum: dummy observation data that implement the prior
% breaks: vector of points in the dummy data after which new dummy obs's start
% Set breaks=T+[0;breaks], ydata=[ydata;ydum], xdum=[xdata;xdum], where
% Set breaks=T+[0;breaks], ydata=[ydata;ydum], xdum=[xdata;xdum], where
% actual data matrix has T rows, in preparing input for rfvar3
% nv,nx,lags: VAR dimensions
% mnprior.tight:Overall tightness of Minnesota prior
@ -175,8 +175,8 @@ function [ydum,xdum,breaks]=varprior(nv,nx,lags,mnprior,vprior)
% taken to include the sum-of-coefficients and co-persistence components
% that are implemented directly in rfvar3.m. The diagonal prior on v, combined
% with sum-of-coefficients and co-persistence components and with the unit own-first-lag
% prior mean generates larger prior variances for own than for cross-effects even in
% this formulation, but here there is no way to shrink toward a set of unconstrained
% prior mean generates larger prior variances for own than for cross-effects even in
% this formulation, but here there is no way to shrink toward a set of unconstrained
% univariate AR's.
% Original file downloaded from:
@ -228,9 +228,9 @@ function var=rfvar3(ydata,lags,xdata,breaks,lambda,mu)
% discontinuities in the data (e.g. war years) and for the possibility of
% adding dummy observations to implement a prior. This must be a column vector.
% Note that a single dummy observation becomes lags+1 rows of the data matrix,
% with a break separating it from the rest of the data. The function treats the
% with a break separating it from the rest of the data. The function treats the
% first lags observations at the top and after each "break" in ydata and xdata as
% initial conditions.
% initial conditions.
% lambda: weight on "co-persistence" prior dummy observations. This expresses
% belief that when data on *all* y's are stable at their initial levels, they will
% tend to persist at that level. lambda=5 is a reasonable first try. With lambda<0,
@ -243,7 +243,7 @@ function var=rfvar3(ydata,lags,xdata,breaks,lambda,mu)
% one of these for each variable. A reasonable first guess is mu=2.
% The program assumes that the first lags rows of ydata and xdata are real data, not dummies.
% Dummy observations should go at the end, if any. If pre-sample x's are not available,
% repeating the initial xdata(lags+1,:) row or copying xdata(lags+1:2*lags,:) into
% repeating the initial xdata(lags+1,:) row or copying xdata(lags+1:2*lags,:) into
% xdata(1:lags,:) are reasonable subsititutes. These values are used in forming the
% persistence priors.
@ -280,7 +280,7 @@ for is = 1:length(smpl)
end
X = [X(:,:) xdata(smpl,:)];
y = ydata(smpl,:);
% Everything now set up with input data for y=Xb+e
% Everything now set up with input data for y=Xb+e
% Add persistence dummies
if lambda ~= 0 || mu > 0

2
matlab/cartesian_product_of_sets.m
View File

@ -5,7 +5,7 @@ function cprod = cartesian_product_of_sets(varargin)
%! @deftypefn {Function File} {@var{cprod} =} cartesian_product_of_sets (@var{a},@var{b}, ...)
%! @anchor{cartesian_product_of_sets}
%! @sp 1
%! Computes A_1 * A_2 * .... * A_n with a generic set A_i = {e_1,e_2,e_3,...} where e_i is a string
%! Computes A_1 * A_2 * .... * A_n with a generic set A_i = {e_1,e_2,e_3,...} where e_i is a string
%! or a number. It is assumed that each element e_i is unique in set A_i.
%! @sp 2
%! @strong{Inputs}

4
matlab/cellofchar2mfile.m
View File

@ -2,11 +2,11 @@ function cellofchar2mfile(fname, c, cname)
% Write a cell of char in a matlab script.
%
% INPUTS
% INPUTS
% - fname [string] name of the file where c is to be saved.
% - c [cell] a two dimensional cell of char.
%
% OUTPUTS
% OUTPUTS
% None.
% Copyright (C) 2015 Dynare Team

5
matlab/check_for_calibrated_covariances.m
View File

@ -42,7 +42,7 @@ calibrated_covariance_pos=covariance_pos(~ismember(covariance_pos,correlation_po
if any(calibrated_covariance_pos)
[rows, columns]=ind2sub(size(M.Sigma_e),calibrated_covariance_pos); %find linear indices of lower triangular covariance entries
estim_params.calibrated_covariances.position=[calibrated_covariance_pos;sub2ind(size(M.Sigma_e),columns,rows)]; %get linear entries of upper triangular parts
estim_params.calibrated_covariances.cov_value=Sigma_e_calibrated(estim_params.calibrated_covariances.position);
estim_params.calibrated_covariances.cov_value=Sigma_e_calibrated(estim_params.calibrated_covariances.position);
end
correlation_pos_ME=find(tril(M.Correlation_matrix_ME,-1)); %off-diagonal elements set by correlations after accounting for estimation
@ -50,6 +50,5 @@ calibrated_covariance_pos_ME=covariance_pos_ME(~ismember(covariance_pos_ME,corre
if any(calibrated_covariance_pos_ME)
[rows, columns]=ind2sub(size(M.H),calibrated_covariance_pos_ME); %find linear indices of lower triangular covariance entries
estim_params.calibrated_covariances_ME.position=[calibrated_covariance_pos_ME;sub2ind(size(M.H),columns,rows)]; %get linear entries of upper triangular parts
estim_params.calibrated_covariances_ME.cov_value=H_calibrated(estim_params.calibrated_covariances_ME.position);
estim_params.calibrated_covariances_ME.cov_value=H_calibrated(estim_params.calibrated_covariances_ME.position);
end

18
matlab/check_list_of_variables.m
View File

@ -1,17 +1,17 @@
function varlist = check_list_of_variables(options_, M_, varlist)
% This function defines, if necessary, the list of endogenous variables
% for which the posterior statistics have to be computed.
%
% for which the posterior statistics have to be computed.
%
% INPUTS
%
% INPUTS
%
% options_ [structure] Dynare structure.
% M_ [structure] Dynare structure (related to model definition).
% varlist [string] Array of strings with name of the endogenous variables.
%
% OUTPUTS
% varlist [string]
%
%
% OUTPUTS
% varlist [string]
%
% SPECIAL REQUIREMENTS
% Copyright (C) 2003-2017 Dynare Team
@ -63,7 +63,7 @@ if ~isempty(varlist) && ~isempty(options_.endo_vars_for_moment_computations_in_e
error('You cannot use the consider_all_endogenous or consider_all_observed options when listing variables after the estimation command')
elseif isempty(varlist) && ~isempty(options_.endo_vars_for_moment_computations_in_estimation)
if strcmp(options_.endo_vars_for_moment_computations_in_estimation,'all_endogenous_variables')
varlist = M_.endo_names(1:M_.orig_endo_nbr, :);
varlist = M_.endo_names(1:M_.orig_endo_nbr, :);
elseif strcmp(options_.endo_vars_for_moment_computations_in_estimation,'only_observed_variables')
varlist = char(options_.varobs');
else
@ -106,7 +106,7 @@ elseif isempty(varlist) && isempty(options_.endo_vars_for_moment_computations_in
end
if ~isempty(cas)
string = [ cas , ' will be computed for the ' num2str(M_.endo_nbr) ' endogenous variables'];
string = [ string ' of your model, this can be very long....'];
string = [ string ' of your model, this can be very long....'];
format_text(string, 10)
if options_.nointeractive
% Default behaviour is to consider all the endogenous variables.

70
matlab/check_matlab_path.m
View File

@ -1,5 +1,5 @@
function check_matlab_path(change_path_flag)
% Copyright (C) 2015-2016 Dynare Team
%
% This file is part of Dynare.
@ -82,7 +82,7 @@ else
warning(msg);
skipline()
rmpath(DYNARE_PATH)
addpath(DYNARE_PATH)
addpath(DYNARE_PATH)
end
warning on backtrace
end
@ -90,7 +90,7 @@ else
else
% Check that the user did not put all the subfolders in the path.
% => If DYNARE_PATH/qz is in the path while mjdgges dll is available
% it most likely means that user wrongly put all subfolders in the
% it most likely means that user wrongly put all subfolders in the
% matlab's path!
mexpath = add_path_to_mex_files([DYNARE_PATH filesep], false);
MATLAB_PATH = path2cell(MATLAB_PATH);
@ -107,40 +107,40 @@ else
end
function q = path2cell(p)
% Converts the output of path() to a cell
s = strfind(p,pathsep);
n = length(s)+1;
q = cell(n,1);
q(1) = {p(1:s(1)-1)};
q(n) = {p(s(end)+1:end)};
for i=2:n-1
q(i) = {p(s(i-1)+1:s(i)-1)};
end
% Converts the output of path() to a cell
s = strfind(p,pathsep);
n = length(s)+1;
q = cell(n,1);
q(1) = {p(1:s(1)-1)};
q(n) = {p(s(end)+1:end)};
for i=2:n-1
q(i) = {p(s(i-1)+1:s(i)-1)};
end
function flist = getallroutinenames(p, excludedsubfolders)
if nargin<2
excludedsubfolders = {};
end
flist={};
%get m-files in this directory
dd = dir([p,filesep '*.m']);
temp=struct2cell(dd);
flist=[flist temp(1,:)];
%deal with subdirectories
dlist=getalldirectories(p,excludedsubfolders); %first call with excluded directories
for ii=1:length(dlist)
flist=[flist getallroutinenames([ p filesep dlist{ii}])]; %recursive calls without subfolders
end
if nargin<2
excludedsubfolders = {};
end
flist={};
%get m-files in this directory
dd = dir([p,filesep '*.m']);
temp=struct2cell(dd);
flist=[flist temp(1,:)];
%deal with subdirectories
dlist=getalldirectories(p,excludedsubfolders); %first call with excluded directories
for ii=1:length(dlist)
flist=[flist getallroutinenames([ p filesep dlist{ii}])]; %recursive calls without subfolders
end
function dlist = getalldirectories(p,excluded_directories)
if nargin<2
excluded_directories = {};
end
dd = dir(p);
dir_result=struct2cell(dd);
directory_indicator=cell2mat(dir_result(4,:));
dlist = dir_result(1,directory_indicator==1 & ~strcmp('.',dir_result(1,:)) & ~strcmp('..',dir_result(1,:)) & ~ismember(dir_result(1,:),excluded_directories));
if nargin<2
excluded_directories = {};
end
dd = dir(p);
dir_result=struct2cell(dd);
directory_indicator=cell2mat(dir_result(4,:));
dlist = dir_result(1,directory_indicator==1 & ~strcmp('.',dir_result(1,:)) & ~strcmp('..',dir_result(1,:)) & ~ismember(dir_result(1,:),excluded_directories));
function n = position(i, currentpath)
n = length(strfind(currentpath(1:i), pathsep));
n = length(strfind(currentpath(1:i), pathsep));

12
matlab/check_posterior_analysis_data.m
View File

@ -1,13 +1,13 @@
function [info,description] = check_posterior_analysis_data(type,M_)
% function [info,description] = check_posterior_analysis_data(type,M_)
% Checks the status of posterior analysis and in particular if files need to be
% Checks the status of posterior analysis and in particular if files need to be
% created or updated; called by posterior_analysis.m
%
%
% Inputs:
% type [string] name of the posterior moment considered
% M_ [structure] Dynare model structure
%
% Outputs:
%
% Outputs:
% info [scalar] return code
% info = 1; % select_posterior_draws has to be called first.
% info = 2; % _posterior_draws files have to be updated.
@ -99,7 +99,7 @@ else
number_of_the_last_post_data_file = length(pdfinfo);
name_of_the_last_post_data_file = ...
[ pwd filesep MetropolisFolder filesep ...
M_.fname '_' ...
M_.fname '_' ...
generic_post_data_file_name ...
int2str(number_of_the_last_post_data_file) ...
'.mat' ];
@ -108,7 +108,7 @@ else
info = 5; % posterior data files have to be updated.
if nargout>1
description = 'posterior data files have to be updated.';
end
end
else
info = 6; % Ok (nothing to do ;-)
if nargout>1

639
matlab/check_posterior_sampler_options.m
View File

@ -40,339 +40,339 @@ if init
% set default options and user defined options
posterior_sampler_options.posterior_sampling_method = options_.posterior_sampler_options.posterior_sampling_method;
posterior_sampler_options.bounds = bounds;
switch posterior_sampler_options.posterior_sampling_method
case 'random_walk_metropolis_hastings'
posterior_sampler_options.parallel_bar_refresh_rate=50;
posterior_sampler_options.serial_bar_refresh_rate=3;
posterior_sampler_options.parallel_bar_title='RWMH';
posterior_sampler_options.serial_bar_title='RW Metropolis-Hastings';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.rwmh);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'proposal_distribution'
if ~(strcmpi(options_list{i,2}, 'rand_multivariate_student') || ...
strcmpi(options_list{i,2}, 'rand_multivariate_normal'))
error(['initial_estimation_checks:: the proposal_distribution option to estimation takes either ' ...
'rand_multivariate_student or rand_multivariate_normal as options']);
else
posterior_sampler_options.proposal_distribution=options_list{i,2};
end
case 'student_degrees_of_freedom'
if options_list{i,2} <= 0
error('initial_estimation_checks:: the student_degrees_of_freedom takes a positive integer argument');
else
posterior_sampler_options.student_degrees_of_freedom=options_list{i,2};
end
case 'use_mh_covariance_matrix'
% indicates to use the covariance matrix from previous iterations to
% define the covariance of the proposal distribution
% default = 0
posterior_sampler_options.use_mh_covariance_matrix = options_list{i,2};
options_.use_mh_covariance_matrix = options_list{i,2};
case 'scale_file'
% load optimal_mh_scale parameter if previous run was with mode_compute=6
% will overwrite jscale from set_prior.m
if exist(options_list{i,2},'file') || exist([options_list{i,2},'.mat'],'file')
tmp = load(options_list{i,2},'Scale');
global bayestopt_
bayestopt_.mh_jscale = tmp.Scale;
options_.mh_jscale = tmp.Scale;
bayestopt_.jscale = ones(size(bounds.lb,1),1)*tmp.Scale;
% options_.mh_init_scale = 2*options_.mh_jscale;
else
error('initial_estimation_checks:: The specified mh_scale_file does not exist.')
end
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['rwmh_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
end
case 'tailored_random_block_metropolis_hastings'
posterior_sampler_options.parallel_bar_refresh_rate=5;
posterior_sampler_options.serial_bar_refresh_rate=1;
posterior_sampler_options.parallel_bar_title='TaRB-MH';
posterior_sampler_options.serial_bar_title='TaRB Metropolis-Hastings';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.tarb);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'proposal_distribution'
if ~(strcmpi(options_list{i,2}, 'rand_multivariate_student') || ...
strcmpi(options_list{i,2}, 'rand_multivariate_normal'))
error(['initial_estimation_checks:: the proposal_distribution option to estimation takes either ' ...
'rand_multivariate_student or rand_multivariate_normal as options']);
else
posterior_sampler_options.proposal_distribution=options_list{i,2};
end
case 'student_degrees_of_freedom'
if options_list{i,2} <= 0
error('initial_estimation_checks:: the student_degrees_of_freedom takes a positive integer argument');
else
posterior_sampler_options.student_degrees_of_freedom=options_list{i,2};
end
case 'mode_compute'
posterior_sampler_options.mode_compute=options_list{i,2};
case 'optim'
posterior_sampler_options.optim_opt=options_list{i,2};
case 'new_block_probability'
if options_list{i,2}<0 || options_list{i,2}>1
error('check_posterior_sampler_options:: The tarb new_block_probability must be between 0 and 1!')
else
posterior_sampler_options.new_block_probability=options_list{i,2};
end
case 'scale_file'
% load optimal_mh_scale parameter if previous run was with mode_compute=6
% will overwrite jscale from set_prior.m
if exist(options_list{i,2},'file') || exist([options_list{i,2},'.mat'],'file')
tmp = load(options_list{i,2},'Scale');
global bayestopt_
bayestopt_.mh_jscale = tmp.Scale;
options_.mh_jscale = tmp.Scale;
bayestopt_.jscale = ones(size(bounds.lb,1),1)*tmp.Scale;
% options_.mh_init_scale = 2*options_.mh_jscale;
else
error('initial_estimation_checks:: The specified scale_file does not exist.')
end
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['tarb_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
end
case 'independent_metropolis_hastings'
posterior_sampler_options.parallel_bar_refresh_rate=50;
posterior_sampler_options.serial_bar_refresh_rate=3;
posterior_sampler_options.parallel_bar_title='IMH';
posterior_sampler_options.serial_bar_title='Ind. Metropolis-Hastings';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.imh);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'proposal_distribution'
if ~(strcmpi(options_list{i,2}, 'rand_multivariate_student') || ...
strcmpi(options_list{i,2}, 'rand_multivariate_normal'))
error(['initial_estimation_checks:: the proposal_distribution option to estimation takes either ' ...
'rand_multivariate_student or rand_multivariate_normal as options']);
else
posterior_sampler_options.proposal_distribution=options_list{i,2};
end
case 'student_degrees_of_freedom'
if options_list{i,2} <= 0
error('initial_estimation_checks:: the student_degrees_of_freedom takes a positive integer argument');
else
posterior_sampler_options.student_degrees_of_freedom=options_list{i,2};
end
case 'use_mh_covariance_matrix'
% indicates to use the covariance matrix from previous iterations to
% define the covariance of the proposal distribution
% default = 0
posterior_sampler_options.use_mh_covariance_matrix = options_list{i,2};
options_.use_mh_covariance_matrix = options_list{i,2};
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
case 'random_walk_metropolis_hastings'
posterior_sampler_options.parallel_bar_refresh_rate=50;
posterior_sampler_options.serial_bar_refresh_rate=3;
posterior_sampler_options.parallel_bar_title='RWMH';
posterior_sampler_options.serial_bar_title='RW Metropolis-Hastings';
otherwise
warning(['imh_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
end
case 'slice'
posterior_sampler_options.parallel_bar_refresh_rate=1;
posterior_sampler_options.serial_bar_refresh_rate=1;
posterior_sampler_options.parallel_bar_title='SLICE';
posterior_sampler_options.serial_bar_title='SLICE';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.slice);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'rotated'
% triggers rotated slice iterations using a covariance
% matrix from initial burn-in iterations
% must be associated with:
% <use_mh_covariance_matrix> or <slice_initialize_with_mode>
% default = 0
posterior_sampler_options.rotated = options_list{i,2};
case 'mode'
% for multimodal posteriors, provide the list of modes as a
% matrix, ordered by column, i.e. [x1 x2 x3] for three
% modes x1 x2 x3
% MR note: not sure this is possible with the
% read_key_value_string ???
% if this is not possible <mode_files> does to job in any case
% This will automatically trigger <rotated>
% default = []
tmp_mode = options_list{i,2};
for j=1:size(tmp_mode,2)
posterior_sampler_options.mode(j).m = tmp_mode(:,j);
end
case 'mode_files'
% for multimodal posteriors provide the name of
% a file containing a variable array xparams = [nparam * nmodes]
% one column per mode. With this info, the code will automatically
% set the <mode> option.
% This will automatically trigger <rotated>
% default = []
posterior_sampler_options.mode_files = options_list{i,2};
case 'slice_initialize_with_mode'
% the default for slice is to set mode_compute = 0 in the
% preprocessor and start the chain(s) from a random location in the prior.
% This option first runs the optimizer and then starts the
% chain from the mode. Associated with optios <rotated>, it will
% use invhess from the mode to perform rotated slice
% iterations.
% default = 0
posterior_sampler_options.slice_initialize_with_mode = options_list{i,2};
case 'initial_step_size'
% sets the initial size of the interval in the STEPPING-OUT PROCEDURE
% the initial_step_size must be a real number in [0, 1],
% and it sets the size as a proportion of the prior bounds,
% i.e. the size will be initial_step_size*(UB-LB)
% slice sampler requires prior_truncation > 0!
% default = 0.8
if options_list{i,2}<=0 || options_list{i,2}>=1
error('check_posterior_sampler_options:: slice initial_step_size must be between 0 and 1')
else
posterior_sampler_options.initial_step_size=options_list{i,2};
end
case 'use_mh_covariance_matrix'
% in association with <rotated> indicates to use the
% covariance matrix from previous iterations to define the
% rotated slice
% default = 0
posterior_sampler_options.use_mh_covariance_matrix = options_list{i,2};
options_.use_mh_covariance_matrix = options_list{i,2};
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.rwmh);
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['slice_sampler: Unknown option (' options_list{i,1} ')!'])
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'proposal_distribution'
if ~(strcmpi(options_list{i,2}, 'rand_multivariate_student') || ...
strcmpi(options_list{i,2}, 'rand_multivariate_normal'))
error(['initial_estimation_checks:: the proposal_distribution option to estimation takes either ' ...
'rand_multivariate_student or rand_multivariate_normal as options']);
else
posterior_sampler_options.proposal_distribution=options_list{i,2};
end
case 'student_degrees_of_freedom'
if options_list{i,2} <= 0
error('initial_estimation_checks:: the student_degrees_of_freedom takes a positive integer argument');
else
posterior_sampler_options.student_degrees_of_freedom=options_list{i,2};
end
case 'use_mh_covariance_matrix'
% indicates to use the covariance matrix from previous iterations to
% define the covariance of the proposal distribution
% default = 0
posterior_sampler_options.use_mh_covariance_matrix = options_list{i,2};
options_.use_mh_covariance_matrix = options_list{i,2};
case 'scale_file'
% load optimal_mh_scale parameter if previous run was with mode_compute=6
% will overwrite jscale from set_prior.m
if exist(options_list{i,2},'file') || exist([options_list{i,2},'.mat'],'file')
tmp = load(options_list{i,2},'Scale');
global bayestopt_
bayestopt_.mh_jscale = tmp.Scale;
options_.mh_jscale = tmp.Scale;
bayestopt_.jscale = ones(size(bounds.lb,1),1)*tmp.Scale;
% options_.mh_init_scale = 2*options_.mh_jscale;
else
error('initial_estimation_checks:: The specified mh_scale_file does not exist.')
end
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['rwmh_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
% slice posterior sampler does not require mode or hessian to run
% needs to be set to 1 to skip parts in dynare_estimation_1.m
% requiring posterior maximization/calibrated smoother before MCMC
options_.mh_posterior_mode_estimation=1;
if ~ posterior_sampler_options.slice_initialize_with_mode
end
case 'tailored_random_block_metropolis_hastings'
posterior_sampler_options.parallel_bar_refresh_rate=5;
posterior_sampler_options.serial_bar_refresh_rate=1;
posterior_sampler_options.parallel_bar_title='TaRB-MH';
posterior_sampler_options.serial_bar_title='TaRB Metropolis-Hastings';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.tarb);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'proposal_distribution'
if ~(strcmpi(options_list{i,2}, 'rand_multivariate_student') || ...
strcmpi(options_list{i,2}, 'rand_multivariate_normal'))
error(['initial_estimation_checks:: the proposal_distribution option to estimation takes either ' ...
'rand_multivariate_student or rand_multivariate_normal as options']);
else
posterior_sampler_options.proposal_distribution=options_list{i,2};
end
case 'student_degrees_of_freedom'
if options_list{i,2} <= 0
error('initial_estimation_checks:: the student_degrees_of_freedom takes a positive integer argument');
else
posterior_sampler_options.student_degrees_of_freedom=options_list{i,2};
end
case 'mode_compute'
posterior_sampler_options.mode_compute=options_list{i,2};
case 'optim'
posterior_sampler_options.optim_opt=options_list{i,2};
case 'new_block_probability'
if options_list{i,2}<0 || options_list{i,2}>1
error('check_posterior_sampler_options:: The tarb new_block_probability must be between 0 and 1!')
else
posterior_sampler_options.new_block_probability=options_list{i,2};
end
case 'scale_file'
% load optimal_mh_scale parameter if previous run was with mode_compute=6
% will overwrite jscale from set_prior.m
if exist(options_list{i,2},'file') || exist([options_list{i,2},'.mat'],'file')
tmp = load(options_list{i,2},'Scale');
global bayestopt_
bayestopt_.mh_jscale = tmp.Scale;
options_.mh_jscale = tmp.Scale;
bayestopt_.jscale = ones(size(bounds.lb,1),1)*tmp.Scale;
% options_.mh_init_scale = 2*options_.mh_jscale;
else
error('initial_estimation_checks:: The specified scale_file does not exist.')
end
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['tarb_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
end
case 'independent_metropolis_hastings'
posterior_sampler_options.parallel_bar_refresh_rate=50;
posterior_sampler_options.serial_bar_refresh_rate=3;
posterior_sampler_options.parallel_bar_title='IMH';
posterior_sampler_options.serial_bar_title='Ind. Metropolis-Hastings';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.imh);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'proposal_distribution'
if ~(strcmpi(options_list{i,2}, 'rand_multivariate_student') || ...
strcmpi(options_list{i,2}, 'rand_multivariate_normal'))
error(['initial_estimation_checks:: the proposal_distribution option to estimation takes either ' ...
'rand_multivariate_student or rand_multivariate_normal as options']);
else
posterior_sampler_options.proposal_distribution=options_list{i,2};
end
case 'student_degrees_of_freedom'
if options_list{i,2} <= 0
error('initial_estimation_checks:: the student_degrees_of_freedom takes a positive integer argument');
else
posterior_sampler_options.student_degrees_of_freedom=options_list{i,2};
end
case 'use_mh_covariance_matrix'
% indicates to use the covariance matrix from previous iterations to
% define the covariance of the proposal distribution
% default = 0
posterior_sampler_options.use_mh_covariance_matrix = options_list{i,2};
options_.use_mh_covariance_matrix = options_list{i,2};
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['imh_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
end
case 'slice'
posterior_sampler_options.parallel_bar_refresh_rate=1;
posterior_sampler_options.serial_bar_refresh_rate=1;
posterior_sampler_options.parallel_bar_title='SLICE';
posterior_sampler_options.serial_bar_title='SLICE';
% default options
posterior_sampler_options = add_fields_(posterior_sampler_options,options_.posterior_sampler_options.slice);
% user defined options
if ~isempty(options_.posterior_sampler_options.sampling_opt)
options_list = read_key_value_string(options_.posterior_sampler_options.sampling_opt);
for i=1:rows(options_list)
switch options_list{i,1}
case 'rotated'
% triggers rotated slice iterations using a covariance
% matrix from initial burn-in iterations
% must be associated with:
% <use_mh_covariance_matrix> or <slice_initialize_with_mode>
% default = 0
posterior_sampler_options.rotated = options_list{i,2};
case 'mode'
% for multimodal posteriors, provide the list of modes as a
% matrix, ordered by column, i.e. [x1 x2 x3] for three
% modes x1 x2 x3
% MR note: not sure this is possible with the
% read_key_value_string ???
% if this is not possible <mode_files> does to job in any case
% This will automatically trigger <rotated>
% default = []
tmp_mode = options_list{i,2};
for j=1:size(tmp_mode,2)
posterior_sampler_options.mode(j).m = tmp_mode(:,j);
end
case 'mode_files'
% for multimodal posteriors provide the name of
% a file containing a variable array xparams = [nparam * nmodes]
% one column per mode. With this info, the code will automatically
% set the <mode> option.
% This will automatically trigger <rotated>
% default = []
posterior_sampler_options.mode_files = options_list{i,2};
case 'slice_initialize_with_mode'
% the default for slice is to set mode_compute = 0 in the
% preprocessor and start the chain(s) from a random location in the prior.
% This option first runs the optimizer and then starts the
% chain from the mode. Associated with optios <rotated>, it will
% use invhess from the mode to perform rotated slice
% iterations.
% default = 0
posterior_sampler_options.slice_initialize_with_mode = options_list{i,2};
case 'initial_step_size'
% sets the initial size of the interval in the STEPPING-OUT PROCEDURE
% the initial_step_size must be a real number in [0, 1],
% and it sets the size as a proportion of the prior bounds,
% i.e. the size will be initial_step_size*(UB-LB)
% slice sampler requires prior_truncation > 0!
% default = 0.8
if options_list{i,2}<=0 || options_list{i,2}>=1
error('check_posterior_sampler_options:: slice initial_step_size must be between 0 and 1')
else
posterior_sampler_options.initial_step_size=options_list{i,2};
end
case 'use_mh_covariance_matrix'
% in association with <rotated> indicates to use the
% covariance matrix from previous iterations to define the
% rotated slice
% default = 0
posterior_sampler_options.use_mh_covariance_matrix = options_list{i,2};
options_.use_mh_covariance_matrix = options_list{i,2};
case 'save_tmp_file'
posterior_sampler_options.save_tmp_file = options_list{i,2};
otherwise
warning(['slice_sampler: Unknown option (' options_list{i,1} ')!'])
end
end
end
% slice posterior sampler does not require mode or hessian to run
% needs to be set to 1 to skip parts in dynare_estimation_1.m
% requiring posterior maximization/calibrated smoother before MCMC
options_.mh_posterior_mode_estimation=1;
if ~ posterior_sampler_options.slice_initialize_with_mode
% by default, slice sampler should trigger
% mode_compute=0 and
% mh_replic=100 (much smaller than the default mh_replic=20000 of RWMH)
options_.mode_compute = 0;
options_.cova_compute = 0;
else
if (isequal(options_.mode_compute,0) && isempty(options_.mode_file) )
skipline()
disp('check_posterior_sampler_options:: You have specified the option "slice_initialize_with_mode"')
disp('check_posterior_sampler_options:: to initialize the slice sampler using mode information')
disp('check_posterior_sampler_options:: but no mode file nor posterior maximization is selected,')
error('check_posterior_sampler_options:: The option "slice_initialize_with_mode" is inconsistent with mode_compute=0 or empty mode_file.')
else
options_.mh_posterior_mode_estimation=0;
end
end
if any(isinf(bounds.lb)) || any(isinf(bounds.ub))
options_.mode_compute = 0;
options_.cova_compute = 0;
else
if (isequal(options_.mode_compute,0) && isempty(options_.mode_file) )
skipline()
disp('some priors are unbounded and prior_trunc is set to zero')
error('The option "slice" is inconsistent with prior_trunc=0.')
end
% moreover slice must be associated to:
% options_.mh_posterior_mode_estimation = 0;
% this is done below, but perhaps preprocessing should do this?
if ~isempty(posterior_sampler_options.mode)
% multimodal case
posterior_sampler_options.rotated = 1;
posterior_sampler_options.WR=[];
end
% posterior_sampler_options = set_default_option(posterior_sampler_options,'mode_files',[]);
posterior_sampler_options.W1=posterior_sampler_options.initial_step_size*(bounds.ub-bounds.lb);
if options_.load_mh_file
posterior_sampler_options.slice_initialize_with_mode = 0;
disp('check_posterior_sampler_options:: You have specified the option "slice_initialize_with_mode"')
disp('check_posterior_sampler_options:: to initialize the slice sampler using mode information')
disp('check_posterior_sampler_options:: but no mode file nor posterior maximization is selected,')
error('check_posterior_sampler_options:: The option "slice_initialize_with_mode" is inconsistent with mode_compute=0 or empty mode_file.')
else
if ~posterior_sampler_options.slice_initialize_with_mode
posterior_sampler_options.invhess=[];
end
options_.mh_posterior_mode_estimation=0;
end
if ~isempty(posterior_sampler_options.mode_files) % multimodal case
modes = posterior_sampler_options.mode_files; % these can be also mean files from previous parallel slice chains
load(modes, 'xparams')
if size(xparams,2)<2
error(['check_posterior_sampler_options:: Variable xparams loaded in file <' modes '> has size [' int2str(size(xparams,1)) 'x' int2str(size(xparams,2)) ']: it must contain at least two columns, to allow multi-modal sampling.'])
end
for j=1:size(xparams,2)
mode(j).m=xparams(:,j);
end
posterior_sampler_options.mode = mode;
posterior_sampler_options.rotated = 1;
posterior_sampler_options.WR=[];
end
if any(isinf(bounds.lb)) || any(isinf(bounds.ub))
skipline()
disp('some priors are unbounded and prior_trunc is set to zero')
error('The option "slice" is inconsistent with prior_trunc=0.')
end
% moreover slice must be associated to:
% options_.mh_posterior_mode_estimation = 0;
% this is done below, but perhaps preprocessing should do this?
if ~isempty(posterior_sampler_options.mode)
% multimodal case
posterior_sampler_options.rotated = 1;
posterior_sampler_options.WR=[];
end
% posterior_sampler_options = set_default_option(posterior_sampler_options,'mode_files',[]);
posterior_sampler_options.W1=posterior_sampler_options.initial_step_size*(bounds.ub-bounds.lb);
if options_.load_mh_file
posterior_sampler_options.slice_initialize_with_mode = 0;
else
if ~posterior_sampler_options.slice_initialize_with_mode
posterior_sampler_options.invhess=[];
end
otherwise
error('check_posterior_sampler_options:: Unknown posterior_sampling_method option %s ',posterior_sampler_options.posterior_sampling_method);
end
if ~isempty(posterior_sampler_options.mode_files) % multimodal case
modes = posterior_sampler_options.mode_files; % these can be also mean files from previous parallel slice chains
load(modes, 'xparams')
if size(xparams,2)<2
error(['check_posterior_sampler_options:: Variable xparams loaded in file <' modes '> has size [' int2str(size(xparams,1)) 'x' int2str(size(xparams,2)) ']: it must contain at least two columns, to allow multi-modal sampling.'])
end
for j=1:size(xparams,2)
mode(j).m=xparams(:,j);
end
posterior_sampler_options.mode = mode;
posterior_sampler_options.rotated = 1;
posterior_sampler_options.WR=[];
end
otherwise
error('check_posterior_sampler_options:: Unknown posterior_sampling_method option %s ',posterior_sampler_options.posterior_sampling_method);
end
return
end
@ -434,6 +434,3 @@ fnam = fieldnames(sampler_options);
for j=1:length(fnam)
posterior_sampler_options.(fnam{j}) = sampler_options.(fnam{j});
end

18
matlab/check_prior_analysis_data.m
View File

@ -1,13 +1,13 @@
function [info,description] = check_prior_analysis_data(type,M_)
% function [info,description] = check_prior_analysis_data(type,M_)
% Checks the status of prior analysis and in particular if files need to be
% Checks the status of prior analysis and in particular if files need to be
% created or updated; called by prior_analysis.m
%
%
% Inputs:
% type [string] name of the posterior moment considered
% M_ [structure] Dynare model structure
%
% Outputs:
%
% Outputs:
% info [scalar] return code
% info = 1; % prior_sampler has to be called first.
% info = 2; % _prior_draws files have to be updated.
@ -40,7 +40,7 @@ if nargout>1
description = '';
end
%% Get informations about prior draws files.
%% Get informations about prior draws files.
if ~exist([ M_.dname '/prior/draws'],'dir')
disp('check_prior_analysis_data:: Can''t find any prior draws file!')
return
@ -65,7 +65,7 @@ else
end
return
else
info = 3; % Nothing to do!
info = 3; % Nothing to do!
if nargout>1
description = 'prior draws files are up to date.';
end
@ -98,9 +98,9 @@ else
name_of_the_last_prior_data_file = pdinfo(end).name;
pdfdate = pdinfo(end).datenum;
% /!\ REMARK /!\
% The user can change the model or the value of a calibrated
% parameter without changing the prior. In this case the (prior)
% moments should be computed. But this case cannot be detected!!!
% The user can change the model or the value of a calibrated
% parameter without changing the prior. In this case the (prior)
% moments should be computed. But this case cannot be detected!!!
if pdfdate<date_of_the_last_prior_draw_file
info = 5; % prior data files have to be updated.
if nargout>1

2
matlab/check_prior_bounds.m
View File

@ -5,7 +5,7 @@ function check_prior_bounds(xparam1,bounds,M_,estim_params_,options_,bayestopt_)
% -xparam1 [double] vector of parameters to be estimated (initial values)
% -bounds [vector] vector containing the lower and upper
% bounds
% -M_ [structure] characterizing the model.
% -M_ [structure] characterizing the model.
% -estim_params_ [structure] characterizing parameters to be estimated
% -options_ [structure] characterizing the options
% -bayestopt_ [structure] characterizing priors

55
matlab/chol_SE.m
View File

@ -1,17 +1,17 @@
function [R,indef, E, P]=chol_SE(A,pivoting)
% [R,indef, E, P]=chol_SE(A,pivoting)
% Performs a (modified) Cholesky factorization of the form
%
% Performs a (modified) Cholesky factorization of the form
%
% P'*A*P + E = R'*R
%
%
% As detailed in Schnabel/Eskow (1990), the factorization has 2 phases:
% Phase 1: While A can still be positive definite, pivot on the maximum diagonal element.
% Check that the standard Cholesky update would result in a positive diagonal
% at the current iteration. If so, continue with the normal Cholesky update.
% Otherwise switch to phase 2.
% If A is safely positive definite, stage 1 is never left, resulting in
% the standard Cholesky decomposition.
%
% at the current iteration. If so, continue with the normal Cholesky update.
% Otherwise switch to phase 2.
% If A is safely positive definite, stage 1 is never left, resulting in
% the standard Cholesky decomposition.
%
% Phase 2: Pivot on the minimum of the negatives of the lower Gershgorin bound
% estimates. To prevent negative diagonals, compute the amount to add to the
% pivot element and add this. Then, do the Cholesky update and update the estimates of the
@ -21,18 +21,18 @@ function [R,indef, E, P]=chol_SE(A,pivoting)
% - During factorization, L=R' is stored in the lower triangle of the original matrix A,
% miminizing the memory requirements
% - Conforming with the original Schnabel/Eskow (1990) algorithm:
% - at each iteration the updated Gershgorin bounds are estimated instead of recomputed,
% - at each iteration the updated Gershgorin bounds are estimated instead of recomputed,
% reducing the computational requirements from o(n^3) to o (n^2)
% - For the last 2 by 2 submatrix, an eigenvalue-based decomposition is used
% - While pivoting is not necessary, it improves the size of E, the add-on on to the diagonal. But this comes at
% - While pivoting is not necessary, it improves the size of E, the add-on on to the diagonal. But this comes at
% the cost of introduding a permutation.
%
%
% Inputs
%
% Inputs
% A [n*n] Matrix to be factorized
% pivoting [scalar] dummy whether pivoting is used
%
% Outputs
%
% Outputs
% R [n*n] originally stored in lower triangular portion of A, including the main diagonal
%
% E [n*1] Elements added to the diagonal of A
@ -40,18 +40,18 @@ function [R,indef, E, P]=chol_SE(A,pivoting)
% performing the decomposition
%
% REFERENCES:
% This implementation is based on
% This implementation is based on
%
% Robert B. Schnabel and Elizabeth Eskow. 1990. "A New Modified Cholesky
% Factorization," SIAM Journal of Scientific Statistical Computating,
% 11, 6: 1136-58.
%
%
% Elizabeth Eskow and Robert B. Schnabel 1991. "Algorithm 695 - Software for a New Modified Cholesky
% Factorization," ACM Transactions on Mathematical Software, Vol 17, No 3: 306-312
%
%
%
%
% Author: Johannes Pfeifer based on Eskow/Schnabel (1991)
%
%
% Copyright (C) 2015 Johannes Pfeifer
% Copyright (C) 2015 Dynare Team
% This file is part of Dynare.
@ -148,7 +148,7 @@ for j = 1:n-1
% Swap elements of the permutation vector
itemp = P(j);
P(j) = P(imaxd);
P(imaxd) = itemp;
P(imaxd) = itemp;
end
end
% check to see whether the normal Cholesky update for this
@ -183,13 +183,13 @@ for j = 1:n-1
g=gersh_nested(A,j,n);
end
end
% PHASE 2
if ~phase1
if j ~= n-1
if pivoting
% Find the minimum negative Gershgorin bound
[tmp,iming] = min(g(j:n));
[tmp,iming] = min(g(j:n));
iming=iming+j-1;
% Pivot to the top the row and column with the
% minimum negative Gershgorin bound
@ -229,9 +229,9 @@ for j = 1:n-1
end
% Calculate delta and add to the diagonal. delta=max{0,-A(j,j) + max{normj,taugam},delta_previous}
% where normj=sum of |A(i,j)|,for i=1,n, delta_previous is the delta computed at the previous iter and taugam is tau1*gamma.
normj=sum(abs(A(j+1:n,j)));
delta = max([0;delta;-A(j,j)+normj;-A(j,j)+taugam]); % get adjustment based on formula on bottom of p. 309 of Eskow/Schnabel (1991)
E(j) = delta;
@ -259,9 +259,9 @@ for j = 1:n-1
% Find eigenvalues of final 2 by 2 submatrix
% Find delta such that:
% 1. the l2 condition number of the final 2X2 submatrix + delta*I <= tau2
% 2. delta >= previous delta,
% 2. delta >= previous delta,
% 3. min(eigvals) + delta >= tau2 * gamma, where min(eigvals) is the smallest eigenvalue of the final 2X2 submatrix
A(n-1,n)=A(n,n-1); %set value above diagonal for computation of eigenvalues
eigvals = eig(A(n-1:n,n-1:n));
delta = max([ 0 ; delta ; -min(eigvals)+tau2*max((max(eigvals)-min(eigvals))/(1-tau1),gammma) ]); %Formula 5.3.2 of Schnabel/Eskow (1990)
@ -272,7 +272,7 @@ for j = 1:n-1
E(n-1) = delta;
E(n) = delta;
end
% Final update
A(n-1,n-1) = sqrt(A(n-1,n-1));
A(n,n-1) = A(n,n-1)/A(n-1,n-1);
@ -306,4 +306,3 @@ for ii = j:n
g(ii) = sum_up_to_i + sum_after_i- A(ii,ii);
end
end

2
matlab/clear_persistent_variables.m
View File

@ -58,5 +58,5 @@ if writelistofroutinestobecleared
return
end
list_of_functions_to_be_cleared;
list_of_functions_to_be_cleared;
clear(list_of_functions{:});

14
matlab/cli/prior.m
View File

@ -42,7 +42,7 @@ if isempty(varargin) || ( isequal(length(varargin), 1) && isequal(varargin{1},'h
return
end
global options_ M_ estim_params_ bayestopt_ oo_
global options_ M_ estim_params_ bayestopt_ oo_
donesomething = false;
@ -53,11 +53,11 @@ end
if (size(estim_params_.var_endo,1) || size(estim_params_.corrn,1))
% Prior over measurement errors are defined...
if ((isfield(options_,'varobs') && isempty(options_.varobs)) || ~isfield(options_,'varobs'))
% ... But the list of observed variabled is not yet defined.
warning('Prior detected on measurement erros, but no list of observed variables (varobs is missing)!')
return
end
if ((isfield(options_,'varobs') && isempty(options_.varobs)) || ~isfield(options_,'varobs'))
% ... But the list of observed variabled is not yet defined.
warning('Prior detected on measurement erros, but no list of observed variables (varobs is missing)!')
return
end
end
% Fill or update bayestopt_ structure
@ -114,7 +114,7 @@ if ismember('optimize', varargin) % Prior optimization.
end
if ismember('moments', varargin) % Prior simulations (2nd order moments).
% Set estimated parameters to the prior mode...
% Set estimated parameters to the prior mode...
xparam1 = BayesOptions.p5;
% ... Except for uniform priors (use the prior mean)!
k = find(isnan(xparam1));

28
matlab/collect_latex_files.m
View File

@ -1,10 +1,10 @@
function collect_latex_files
% function collect_LaTeX_Files;
% Creates TeX-File embedding all eps-loaders created for current mod-file
%
%
% Inputs: none
%
% Notes:
%
% Notes:
% - The packages loaded enable pdflatex to run
% - The _dynamic and _static TeX-model files are not included as they are standalone TeX-files
@ -45,11 +45,11 @@ TeX_Files=dir([M_.fname,'*.tex']);
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_dynamic.tex']) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_static.tex']) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_original.tex']) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_TeX_binder.tex'])
fprintf(fid,'%s \n',['\include{',f_name,'}']);
~strcmp(TeX_Files(ii).name,[M_.fname,'_dynamic.tex']) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_static.tex']) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_original.tex']) && ...
~strcmp(TeX_Files(ii).name,[M_.fname,'_TeX_binder.tex'])
fprintf(fid,'%s \n',['\include{',f_name,'}']);
end
end
@ -58,7 +58,7 @@ TeX_Files=dir([M_.dname filesep 'Output' filesep M_.fname '*.tex']);
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder)
fprintf(fid,'%s \n',['\include{', M_.dname '/Output' '/',f_name,'}']);
fprintf(fid,'%s \n',['\include{', M_.dname '/Output' '/',f_name,'}']);
end
end
@ -67,7 +67,7 @@ TeX_Files=dir([M_.dname filesep 'graphs' filesep M_.fname '*.tex']);
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder)
fprintf(fid,'%s \n',['\include{', M_.dname '/graphs' '/',f_name,'}']);
fprintf(fid,'%s \n',['\include{', M_.dname '/graphs' '/',f_name,'}']);
end
end
@ -76,7 +76,7 @@ TeX_Files=dir([M_.dname filesep 'identification' filesep M_.fname '*.tex']);
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder)
fprintf(fid,'%s \n',['\include{', M_.dname '/identification' '/',f_name,'}']);
fprintf(fid,'%s \n',['\include{', M_.dname '/identification' '/',f_name,'}']);
end
end
@ -86,7 +86,7 @@ TeX_Files=dir([M_.dname filesep 'identification' filesep 'Output' filesep M_.fna
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder)
fprintf(fid,'%s \n',['\include{', M_.dname '/identification/Output' '/',f_name,'}']);
fprintf(fid,'%s \n',['\include{', M_.dname '/identification/Output' '/',f_name,'}']);
end
end
@ -95,7 +95,7 @@ TeX_Files=dir([M_.dname filesep 'gsa' filesep M_.fname '*.tex']);
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder)
fprintf(fid,'%s \n',['\include{', M_.dname '/gsa' '/',f_name,'}']);
fprintf(fid,'%s \n',['\include{', M_.dname '/gsa' '/',f_name,'}']);
end
end
@ -104,7 +104,7 @@ TeX_Files=dir([M_.dname filesep 'gsa' filesep 'Output' filesep M_.fname '*.tex'
for ii=1:length(TeX_Files)
[pathstr,f_name,ext] = fileparts(TeX_Files(ii).name);
if ~strcmp(TeX_Files(ii).name,f_name_binder)
fprintf(fid,'%s \n',['\include{', M_.dname '/gsa/Output' '/',f_name,'}']);
fprintf(fid,'%s \n',['\include{', M_.dname '/gsa/Output' '/',f_name,'}']);
end
end

23
matlab/compute_Pinf_Pstar.m
View File

@ -1,30 +1,30 @@
function [Pstar,Pinf] = compute_Pinf_Pstar(mf,T,R,Q,qz_criterium, restrict_columns)
% function [Z,ST,QT,R1,Pstar,Pinf] = schur_statespace_transformation(mf,T,R,Q,qz_criterium, restrict_columns)
% Kitagawa transformation of state space system with a quasi-triangular
% transition matrix with unit roots at the top, but excluding zero columns of the transition matrix.
% transition matrix with unit roots at the top, but excluding zero columns of the transition matrix.
% Computation of Pstar and Pinf for Durbin and Koopman Diffuse filter
%
% The transformed state space is
% The transformed state space is
% y = [ss; z; x];
% s = static variables (zero columns of T)
% z = unit roots
% x = stable roots
% ss = s - z = stationarized static variables
%
% INPUTS
%
% INPUTS
% mf [integer] vector of indices of observed variables in
% state vector
% T [double] matrix of transition
% R [double] matrix of structural shock effects
% Q [double] matrix of covariance of structural shocks
% qz_criterium [double] numerical criterium for unit roots
%
% OUTPUTS
% qz_criterium [double] numerical criterium for unit roots
%
% OUTPUTS
% Pstar [double] matrix of covariance of stationary part
% Pinf [double] matrix of covariance initialization for
% nonstationary part
%
% ALGORITHM
% nonstationary part
%
% ALGORITHM
% Real Schur transformation of transition equation
%
% SPECIAL REQUIREMENTS
@ -49,7 +49,7 @@ function [Pstar,Pinf] = compute_Pinf_Pstar(mf,T,R,Q,qz_criterium, restrict_colum
np = size(T,1);
% perform Kitagawa transformation
% perform Kitagawa transformation
[QT,ST] = schur(T);
e1 = abs(ordeig(ST)) > 2-qz_criterium;
[QT,ST] = ordschur(QT,ST,e1);
@ -112,4 +112,3 @@ end
Pinf = QT*Pinf*QT';
Pstar = QT*Pstar*QT';

12
matlab/compute_mh_covariance_matrix.m
View File

@ -1,15 +1,15 @@
function [posterior_mean,posterior_covariance,posterior_mode,posterior_kernel_at_the_mode] = compute_mh_covariance_matrix()
% Estimation of the posterior covariance matrix, posterior mean, posterior mode and evaluation of the posterior kernel at the
% estimated mode, using the draws from a metropolis-hastings. The estimated posterior mode and covariance matrix are saved in
% a file <M_.fname>_mh_mode.mat.
%
% INPUTS
% a file <M_.fname>_mh_mode.mat.
%
% INPUTS
% None.
%
%
% OUTPUTS
% o posterior_mean [double] (n*1) vector, posterior expectation of the parameters.
% o posterior_covariance [double] (n*n) matrix, posterior covariance of the parameters (computed from previous metropolis hastings).
% o posterior_mode [double] (n*1) vector, posterior mode of the parameters.
% o posterior_mode [double] (n*1) vector, posterior mode of the parameters.
% o posterior_kernel_at_the_mode [double] scalar.
%
% SPECIAL REQUIREMENTS
@ -62,7 +62,7 @@ offset = 0;
for b=1:nblck
first_line = FirstLine;
for n = FirstMhFile:TotalNumberOfMhFiles
load([ BaseName '_mh' int2str(n) '_blck' int2str(b) '.mat'],'x2','logpo2');
load([ BaseName '_mh' int2str(n) '_blck' int2str(b) '.mat'],'x2','logpo2');
[tmp,idx] = max(logpo2);
if tmp>posterior_kernel_at_the_mode
posterior_kernel_at_the_mode = tmp;

4
matlab/compute_model_moments.m
View File

@ -4,10 +4,10 @@ function moments=compute_model_moments(dr,M_,options_)
% dr: structure describing model solution
% M_: structure of Dynare options
% options_
%
%
% OUTPUTS
% moments: a cell array containing requested moments
%
%
% SPECIAL REQUIREMENTS
% none

10
matlab/compute_moments_varendo.m
View File

@ -2,14 +2,14 @@ function oo_ = compute_moments_varendo(type,options_,M_,oo_,var_list_)
% Computes the second order moments (autocorrelation function, covariance
% matrix and variance decomposition) distributions for all the endogenous variables selected in
% var_list_. The results are saved in oo_
%
%
% INPUTS:
% type [string] 'posterior' or 'prior'
% options_ [structure] Dynare structure.
% M_ [structure] Dynare structure (related to model definition).
% oo_ [structure] Dynare structure (results).
% var_list_ [string] Array of string with endogenous variable names.
%
%
% OUTPUTS
% oo_ [structure] Dynare structure (results).
%
@ -31,7 +31,7 @@ function oo_ = compute_moments_varendo(type,options_,M_,oo_,var_list_)
% 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/>.
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
fprintf('Estimation::compute_moments_varendo: I''m computing endogenous moments (this may take a while)... ');
@ -129,7 +129,7 @@ if M_.exo_nbr > 1
headers = char(' ',headers);
lh = size(deblank(var_list_),2)+2;
dyntable(options_,title,headers,deblank(var_list_),100* ...
temp,lh,8,2);
temp,lh,8,2);
if options_.TeX
headers=M_.exo_names_tex;
headers = char(' ',headers);
@ -166,7 +166,7 @@ if M_.exo_nbr > 1
headers = char(' ',headers);
lh = size(deblank(var_list_),2)+2;
dyntable(options_,title_print,headers,deblank(var_list_),100* ...
temp(:,:,step_iter),lh,8,2);
temp(:,:,step_iter),lh,8,2);
if options_.TeX
headers=M_.exo_names_tex;
headers = char(' ',headers);

6
matlab/compute_overall_acceptance_ratio.m
View File

@ -16,7 +16,7 @@ function overallacceptanceratio = compute_overall_acceptance_ratio(MetropolisFol
%
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
BaseName = [MetropolisFolder filesep ModelName];
mh_history_files = dir([BaseName '_mh_history_*.mat']);
@ -24,12 +24,12 @@ n = length(mh_history_files);
load([BaseName '_mh_history_' num2str(0)]);
TotalNumberOfDraws = record.MhDraws(end,1);
TotalNumberOfAcceptedProposals = record.AcceptanceRatio*record.MhDraws(end,1);
TotalNumberOfAcceptedProposals = record.AcceptanceRatio*record.MhDraws(end,1);
for i=2:n
load([BaseName '_mh_history_' num2str(i-1)]);
TotalNumberOfDraws = TotalNumberOfDraws + record.MhDraws(end,1);
TotalNumberOfAcceptedProposals = TotalNumberOfAcceptedProposals + record.AcceptanceRatio*record.MhDraws(end,1);
TotalNumberOfAcceptedProposals = TotalNumberOfAcceptedProposals + record.AcceptanceRatio*record.MhDraws(end,1);
end
overallacceptanceratio = TotalNumberOfAcceptedProposals/TotalNumberOfDraws;

12
matlab/conditional_variance_decomposition.m
View File

@ -1,14 +1,14 @@
function ConditionalVarianceDecomposition = conditional_variance_decomposition(StateSpaceModel, Steps, SubsetOfVariables,sigma_e_is_diagonal)
% This function computes the conditional variance decomposition of a given state space model
% for a subset of endogenous variables.
%
% INPUTS
%
% INPUTS
% StateSpaceModel [structure] Specification of the state space model.
% Steps [integer] 1*h vector of dates.
% SubsetOfVariables [integer] 1*q vector of indices.
%
% OUTPUTS
% ConditionalVarianceDecomposition [double] [n h p] array, where
%
% OUTPUTS
% ConditionalVarianceDecomposition [double] [n h p] array, where
% n is equal to length(SubsetOfVariables)
% h is the number of Steps
% p is the number of state innovations and
@ -77,7 +77,7 @@ for h = 1:length(Steps)
SumOfVariances(:,h) = sum(ConditionalVariance(:,h,:),3);
end
ConditionalVarianceDecomposition = zeros(NumberOfVariables,length(Steps),number_of_state_innovations);
ConditionalVarianceDecomposition = zeros(NumberOfVariables,length(Steps),number_of_state_innovations);
for i=1:number_of_state_innovations
for h = 1:length(Steps)
ConditionalVarianceDecomposition(:,h,i) = squeeze(ConditionalVariance(:,h,i))./SumOfVariances(:,h);

4
matlab/conditional_variance_decomposition_mc_analysis.m
View File

@ -9,7 +9,7 @@ function oo_ = ...
% dname [string] directory name where to save
% fname [string] name of the mod-file
% Steps [integers] horizons at which to conduct decomposition
% exonames [string] (n_exo*char_length) character array with names of exogenous variables
% exonames [string] (n_exo*char_length) character array with names of exogenous variables
% exo [string] name of current exogenous
% variable
% var_list [string] (n_endo*char_length) character array with name
@ -106,7 +106,7 @@ for i=1:length(Steps)
p_density(:,:,i) = pp_density;
else
[pp_mean, pp_median, pp_var, hpd_interval, pp_deciles] = ...
posterior_moments(tmp(:,i),0,mh_conf_sig);
posterior_moments(tmp(:,i),0,mh_conf_sig);
end
p_mean(i) = pp_mean;
p_median(i) = pp_median;

77
matlab/convergence_diagnostics/McMCDiagnostics.m
View File

@ -1,14 +1,14 @@
function oo_ = McMCDiagnostics(options_, estim_params_, M_, oo_)
% function McMCDiagnostics
% Computes convergence tests
%
% INPUTS
% Computes convergence tests
%
% INPUTS
% options_ [structure]
% estim_params_ [structure]
% M_ [structure]
%
% OUTPUTS
% oo_ [structure]
% OUTPUTS
% oo_ [structure]
%
% SPECIAL REQUIREMENTS
% none
@ -82,12 +82,12 @@ param_name_tex=[];
for jj=1:npar
if options_.TeX
[par_name_temp,par_name_tex_temp]=get_the_name(jj,options_.TeX,M_,estim_params_,options_);
param_name = strvcat(param_name,par_name_temp);
param_name = strvcat(param_name,par_name_temp);
par_name_tex_temp = strrep(par_name_tex_temp,'$','');
param_name_tex = strvcat(param_name_tex,par_name_tex_temp);
else
[par_name_temp]=get_the_name(jj,options_.TeX,M_,estim_params_,options_);
param_name = strvcat(param_name,par_name_temp);
param_name = strvcat(param_name,par_name_temp);
end
Draws = GetAllPosteriorDraws(jj,FirstMhFile,FirstLine,TotalNumberOfMhFiles,NumberOfDraws);
Draws = reshape(Draws,[NumberOfDraws nblck]);
@ -130,7 +130,7 @@ end
if nblck == 1 % Brooks and Gelman tests need more than one block
convergence_diagnostics_geweke=zeros(npar,4+2*length(options_.convergence.geweke.taper_steps));
if any(options_.convergence.geweke.geweke_interval<0) || any(options_.convergence.geweke.geweke_interval>1) || length(options_.convergence.geweke.geweke_interval)~=2 ...
|| (options_.convergence.geweke.geweke_interval(2)-options_.convergence.geweke.geweke_interval(1)<0)
|| (options_.convergence.geweke.geweke_interval(2)-options_.convergence.geweke.geweke_interval(1)<0)
fprintf('\nCONVERGENCE DIAGNOSTICS: Invalid option for geweke_interval. Using the default of [0.2 0.5].\n')
options_.convergence.geweke.geweke_interval=[0.2 0.5];
end
@ -139,9 +139,9 @@ if nblck == 1 % Brooks and Gelman tests need more than one block
first_obs_end_sample = first_obs_begin_sample+round(options_.convergence.geweke.geweke_interval(2)*NumberOfDraws*(1-options_.mh_drop));
param_name=[];
if options_.TeX
param_name_tex=[];
param_name_tex=[];
end
for jj=1:npar
for jj=1:npar
if options_.TeX
[param_name_temp, param_name_tex_temp]= get_the_name(jj,options_.TeX,M_,estim_params_,options_);
param_name_tex = strvcat(param_name_tex,strrep(param_name_tex_temp,'$',''));
@ -152,7 +152,7 @@ if nblck == 1 % Brooks and Gelman tests need more than one block
end
end
fprintf('\nGeweke (1992) Convergence Tests, based on means of draws %d to %d vs %d to %d.\n',first_obs_begin_sample,last_obs_begin_sample,first_obs_end_sample,NumberOfDraws);
fprintf('p-values are for Chi2-test for equality of means.\n');
fprintf('p-values are for Chi2-test for equality of means.\n');
Geweke_header=char('Parameter', 'Post. Mean', 'Post. Std', 'p-val No Taper');
for ii=1:length(options_.convergence.geweke.taper_steps)
Geweke_header=char(Geweke_header,['p-val ' num2str(options_.convergence.geweke.taper_steps(ii)),'% Taper']);
@ -168,12 +168,12 @@ if nblck == 1 % Brooks and Gelman tests need more than one block
end
[results_vec, results_struct] = geweke_moments(param_draws,options_);
convergence_diagnostics_geweke(jj,:)=results_vec;
param_draws1 = param_draws(first_obs_begin_sample:last_obs_begin_sample,:);
param_draws2 = param_draws(first_obs_end_sample:end,:);
[results_vec1] = geweke_moments(param_draws1,options_);
[results_vec2] = geweke_moments(param_draws2,options_);
results_struct = geweke_chi2_test(results_vec1,results_vec2,results_struct,options_);
eval(['oo_.convergence.geweke.',param_name(jj,:),'=results_struct;'])
datamat(jj,:)=[results_struct.posteriormean,results_struct.posteriorstd,results_struct.prob_chi2_test];
@ -183,29 +183,29 @@ if nblck == 1 % Brooks and Gelman tests need more than one block
if options_.TeX
Geweke_tex_header=char('Parameter', 'Mean', 'Std', 'No\ Taper');
additional_header={[' & \multicolumn{2}{c}{Posterior} & \multicolumn{',num2str(1+length(options_.convergence.geweke.taper_steps)),'}{c}{p-values} \\'],
['\cmidrule(r{.75em}){2-3} \cmidrule(r{.75em}){4-',num2str(4+length(options_.convergence.geweke.taper_steps)),'}']};
['\cmidrule(r{.75em}){2-3} \cmidrule(r{.75em}){4-',num2str(4+length(options_.convergence.geweke.taper_steps)),'}']};
for ii=1:length(options_.convergence.geweke.taper_steps)
Geweke_tex_header=char(Geweke_tex_header,[num2str(options_.convergence.geweke.taper_steps(ii)),'\%%\ Taper']);
end
headers = char(Geweke_tex_header);
lh = size(param_name_tex,2)+2;
my_title=sprintf('Geweke (1992) Convergence Tests, based on means of draws %d to %d vs %d to %d. p-values are for $\\\\chi^2$-test for equality of means.',first_obs_begin_sample,last_obs_begin_sample,first_obs_end_sample,NumberOfDraws);
dyn_latex_table(M_,options_,my_title,'geweke',headers,param_name_tex,datamat,lh,12,4,additional_header);
dyn_latex_table(M_,options_,my_title,'geweke',headers,param_name_tex,datamat,lh,12,4,additional_header);
end
skipline(2);
if options_.convergence.rafterylewis.indicator
if any(options_.convergence.rafterylewis.qrs<0) || any(options_.convergence.rafterylewis.qrs>1) || length(options_.convergence.rafterylewis.qrs)~=3 ...
|| (options_.convergence.rafterylewis.qrs(1)-options_.convergence.rafterylewis.qrs(2)<=0)
|| (options_.convergence.rafterylewis.qrs(1)-options_.convergence.rafterylewis.qrs(2)<=0)
fprintf('\nCONVERGENCE DIAGNOSTICS: Invalid option for raftery_lewis_qrs. Using the default of [0.025 0.005 0.95].\n')
options_.convergence.rafterylewis.qrs=[0.025 0.005 0.95];
end
end
Raftery_Lewis_q=options_.convergence.rafterylewis.qrs(1);
Raftery_Lewis_r=options_.convergence.rafterylewis.qrs(2);
Raftery_Lewis_s=options_.convergence.rafterylewis.qrs(3);
oo_.Raftery_Lewis = raftery_lewis(x2,Raftery_Lewis_q,Raftery_Lewis_r,Raftery_Lewis_s);
oo_.Raftery_Lewis.parameter_names=param_name;
my_title=sprintf('Raftery/Lewis (1992) Convergence Diagnostics, based on quantile q=%4.3f with precision r=%4.3f with probability s=%4.3f.',Raftery_Lewis_q,Raftery_Lewis_r,Raftery_Lewis_s);
my_title=sprintf('Raftery/Lewis (1992) Convergence Diagnostics, based on quantile q=%4.3f with precision r=%4.3f with probability s=%4.3f.',Raftery_Lewis_q,Raftery_Lewis_r,Raftery_Lewis_s);
headers = char('Variables','M (burn-in)','N (req. draws)','N+M (total draws)','k (thinning)');
raftery_data_mat=[oo_.Raftery_Lewis.M_burn,oo_.Raftery_Lewis.N_prec,oo_.Raftery_Lewis.N_total,oo_.Raftery_Lewis.k_thin];
@ -217,15 +217,15 @@ if nblck == 1 % Brooks and Gelman tests need more than one block
labels_Raftery_Lewis_tex=char(param_name_tex,'Maximum');
lh = size(labels_Raftery_Lewis_tex,2)+2;
dyn_latex_table(M_,options_,my_title,'raftery_lewis',headers,labels_Raftery_Lewis_tex,raftery_data_mat,lh,10,0);
end
end
end
return
end
Origin = 1000;
StepSize = ceil((NumberOfDraws-Origin)/100);% So that the computational time does not
ALPHA = 0.2; % increase too much with the number of simulations.
StepSize = ceil((NumberOfDraws-Origin)/100);% So that the computational time does not
ALPHA = 0.2; % increase too much with the number of simulations.
time = 1:NumberOfDraws;
xx = Origin:StepSize:NumberOfDraws;
NumberOfLines = length(xx);
@ -273,7 +273,7 @@ else
ModelName = [ModelName '_bvar'];
end
NamFileInput={[M_.dname '/metropolis/'],[ModelName '_mh*_blck*.mat']};
[fout, nBlockPerCPU, totCPU] = masterParallel(options_.parallel, 1, npar,NamFileInput,'McMCDiagnostics_core', localVars, [], options_.parallel_info);
UDIAG = fout(1).UDIAG;
for j=2:totCPU
@ -283,9 +283,9 @@ end
UDIAG(:,[2 4 6],:) = UDIAG(:,[2 4 6],:)/nblck;
skipline()
clear pmet temp moyenne CSUP CINF csup cinf n linea iter tmp;
clear pmet temp moyenne CSUP CINF csup cinf n linea iter tmp;
pages = floor(npar/3);
k = 0;
k = 0;
for i = 1:pages
h=dyn_figure(options_.nodisplay,'Name','MCMC univariate convergence diagnostic (Brooks and Gelman,1998)');
boxplot = 1;
@ -296,12 +296,12 @@ for i = 1:pages
if crit == 1
plt1 = UDIAG(:,1,k);
plt2 = UDIAG(:,2,k);
namnam = [nam , ' (Interval)'];
namnam = [nam , ' (Interval)'];
elseif crit == 2
plt1 = UDIAG(:,3,k);
plt2 = UDIAG(:,4,k);
namnam = [nam , ' (m2)'];
elseif crit == 3
elseif crit == 3
plt1 = UDIAG(:,5,k);
plt2 = UDIAG(:,6,k);
namnam = [nam , ' (m3)'];
@ -330,7 +330,7 @@ for i = 1:pages
fprintf(fidTeX,'\\begin{figure}[H]\n');
for jj = 1:size(NAMES,1)
fprintf(fidTeX,'\\psfrag{%s}[1][][0.5][0]{%s}\n',deblank(NAMES(jj,:)),deblank(TEXNAMES(jj,:)));
end
end
fprintf(fidTeX,'\\centering \n');
fprintf(fidTeX,'\\includegraphics[width=%2.2f\\textwidth]{%s_udiag%s}\n',options_.figures.textwidth*min((boxplot-1)/3,1),[OutputFolder '/' ModelName],int2str(i));
fprintf(fidTeX,'\\caption{Univariate convergence diagnostics for the Metropolis-Hastings.\n');
@ -359,12 +359,12 @@ if reste
if crit == 1
plt1 = UDIAG(:,1,k);
plt2 = UDIAG(:,2,k);
namnam = [nam , ' (Interval)'];
namnam = [nam , ' (Interval)'];
elseif crit == 2
plt1 = UDIAG(:,3,k);
plt2 = UDIAG(:,4,k);
namnam = [nam , ' (m2)'];
elseif crit == 3
elseif crit == 3
plt1 = UDIAG(:,5,k);
plt2 = UDIAG(:,6,k);
namnam = [nam , ' (m3)'];
@ -393,7 +393,7 @@ if reste
fprintf(fidTeX,'\\begin{figure}[H]\n');
for jj = 1:size(NAMES,1)
fprintf(fidTeX,'\\psfrag{%s}[1][][0.5][0]{%s}\n',deblank(NAMES(jj,:)),deblank(TEXNAMES(jj,:)));
end
end
fprintf(fidTeX,'\\centering \n');
fprintf(fidTeX,'\\includegraphics[width=%2.2f\\textwidth]{%s_udiag%s}\n',options_.figures.textwidth*min((boxplot-1)/nc,1),[OutputFolder '/' ModelName],int2str(pages+1));
if reste == 2
@ -435,7 +435,7 @@ for b = 1:nblck
end
clear logpo2;
tmp(:,2) = kron(transpose(1:nblck),ones(NumberOfDraws,1));
tmp(:,3) = kron(ones(nblck,1),time');
tmp(:,3) = kron(ones(nblck,1),time');
tmp = sortrows(tmp,1);
ligne = 0;
for iter = Origin:StepSize:NumberOfDraws
@ -454,12 +454,12 @@ for iter = Origin:StepSize:NumberOfDraws
for i=1:nblck
pmet = temp(find(temp(:,2)==i));
MDIAG(ligne,2) = MDIAG(ligne,2) + pmet(csup,1)-pmet(cinf,1);
moyenne = mean(pmet,1); %% Within mean.
moyenne = mean(pmet,1); %% Within mean.
MDIAG(ligne,4) = MDIAG(ligne,4) + sum((pmet(:,1)-moyenne).^2)/(n-1);
MDIAG(ligne,6) = MDIAG(ligne,6) + sum(abs(pmet(:,1)-moyenne).^3)/(n-1);
end
end
MDIAG(:,[2 4 6],:) = MDIAG(:,[2 4 6],:)/nblck;
MDIAG(:,[2 4 6],:) = MDIAG(:,[2 4 6],:)/nblck;
h = dyn_figure(options_.nodisplay,'Name','Multivariate convergence diagnostic');
boxplot = 1;
@ -467,12 +467,12 @@ for crit = 1:3
if crit == 1
plt1 = MDIAG(:,1);
plt2 = MDIAG(:,2);
namnam = 'Interval';
namnam = 'Interval';
elseif crit == 2
plt1 = MDIAG(:,3);
plt2 = MDIAG(:,4);
namnam = 'm2';
elseif crit == 3
elseif crit == 3
plt1 = MDIAG(:,5);
plt2 = MDIAG(:,6);
namnam = 'm3';
@ -499,7 +499,7 @@ if TeX && any(strcmp('eps',cellstr(options_.graph_format)))
fprintf(fidTeX,'\\begin{figure}[H]\n');
for jj = 1:3
fprintf(fidTeX,'\\psfrag{%s}[1][][0.5][0]{%s}\n',deblank(NAMES(jj,:)),' ');
end
end
fprintf(fidTeX,'\\centering \n');
fprintf(fidTeX,'\\includegraphics[width=0.8\\textwidth]{%s_mdiag}\n',[OutputFolder '/' ModelName]);
fprintf(fidTeX,'\\caption{Multivariate convergence diagnostics for the Metropolis-Hastings.\n');
@ -512,4 +512,3 @@ if TeX && any(strcmp('eps',cellstr(options_.graph_format)))
fprintf(fidTeX,'%% End Of TeX file.');
fclose(fidTeX);
end

12
matlab/convergence_diagnostics/McMCDiagnostics_core.m
View File

@ -2,12 +2,12 @@ function myoutput = McMCDiagnostics_core(myinputs,fpar,npar,whoiam, ThisMatlab)
% function myoutput = McMCDiagnostics_core(myinputs,fpar,npar,whoiam, ThisMatlab)
% Computes the Brooks/Gelman (1998) convergence diagnostics, both the
% parameteric and the non-parameteric versions
%
%
% PARALLEL CONTEXT
% Core functionality for MCMC Diagnostics, which can be parallelized.
% See also the comment in posterior_sampler_core.m funtion.
%
%
%
%
% INPUTS
% See See the comment in posterior_sampler_core.m funtion.
@ -25,10 +25,10 @@ function myoutput = McMCDiagnostics_core(myinputs,fpar,npar,whoiam, ThisMatlab)
%
% ALGORITHM
% Computes part of the convergence diagnostics, the rest is computed in McMCDiagnostics.m .
% The methodology and terminology is based on: Brooks/Gelman (1998): General
% Methods for Monitoring Convergence of Iterative Simulations, Journal of Computational
% The methodology and terminology is based on: Brooks/Gelman (1998): General
% Methods for Monitoring Convergence of Iterative Simulations, Journal of Computational
% and Graphical Statistics, Volume 7, Number 4, Pages 434-455
%
%
%
% SPECIAL REQUIREMENTS.
% None.

35
matlab/convergence_diagnostics/geweke_chi2_test.m
View File

@ -2,7 +2,7 @@ function results_struct = geweke_chi2_test(results1,results2,results_struct,opti
% results_struct = geweke_chi2_test(results1,results2,results_struct,options)
% PURPOSE: computes Geweke's chi-squared test for two sets of MCMC sample draws
%
% INPUTS
% INPUTS
% results1 [1 by (4+n_taper*2) vector] vector with post. mean,
% std, NSE_iid, RNE_iid, and tapered NSE and RNE
% for chain part 1
@ -12,7 +12,7 @@ function results_struct = geweke_chi2_test(results1,results2,results_struct,opti
% results_struct [structure] results structure generated by geweke_moments
% Dynareoptions [structure]
%
% OUTPUTS
% OUTPUTS
% results_struct [structure] containing the following fields:
% pooled_mean Pooled mean of the chain parts, weighted
% with precision
@ -49,26 +49,25 @@ function results_struct = geweke_chi2_test(results1,results2,results_struct,opti
% J.M. Bernardo, A.P. Dawid, and A.F.M. Smith (eds.) Proceedings of
% the Fourth Valencia International Meeting on Bayesian Statistics,
% pp. 169-194, Oxford University Press
% Geweke (1999): `Using simulation methods for Bayesian econometric models:
% Geweke (1999): `Using simulation methods for Bayesian econometric models:
% Inference, development and communication', Econometric Reviews, 18(1),
% 1-73
% written by: Johannes Pfeifer,
% based on code by James P. LeSage, who in turn
% drew on MATLAB programs written by Siddartha Chib
% written by: Johannes Pfeifer,
% based on code by James P. LeSage, who in turn
% drew on MATLAB programs written by Siddartha Chib
for k=1:length(options.convergence.geweke.taper_steps)+1
NSE=[results1(:,3+(k-1)*2) results2(:,3+(k-1)*2)];
means=[results1(:,1) results2(:,1)];
diff_Means=means(:,1)-means(:,2);
sum_of_weights=sum(1./(NSE.^2),2);
pooled_mean=sum(means./(NSE.^2),2)./sum_of_weights;
pooled_NSE=1./sqrt(sum_of_weights);
NSE=[results1(:,3+(k-1)*2) results2(:,3+(k-1)*2)];
means=[results1(:,1) results2(:,1)];
diff_Means=means(:,1)-means(:,2);
sum_of_weights=sum(1./(NSE.^2),2);
pooled_mean=sum(means./(NSE.^2),2)./sum_of_weights;
pooled_NSE=1./sqrt(sum_of_weights);
test_stat=diff_Means.^2./sum(NSE.^2,2);
p = 1-chi2cdf(test_stat,1);
results_struct.pooled_mean(:,k) = pooled_mean;
results_struct.pooled_nse(:,k) = pooled_NSE;
results_struct.prob_chi2_test(:,k) = p;
test_stat=diff_Means.^2./sum(NSE.^2,2);
p = 1-chi2cdf(test_stat,1);
results_struct.pooled_mean(:,k) = pooled_mean;
results_struct.pooled_nse(:,k) = pooled_NSE;
results_struct.prob_chi2_test(:,k) = p;
end

23
matlab/convergence_diagnostics/geweke_moments.m
View File

@ -1,13 +1,13 @@
function [results_vec, results_struct] = geweke_moments(draws,Dynareoptions)
%[results_vec, results_struct] = geweke_moments(draws,Dynareoptions)
% PURPOSE: computes Gewke's convergence diagnostics NSE and RNE
% PURPOSE: computes Gewke's convergence diagnostics NSE and RNE
% (numerical std error and relative numerical efficiencies)
% INPUTS
% draws [ndraws by 1 vector]
% INPUTS
% draws [ndraws by 1 vector]
% Dynareoptions [structure]
%
% OUTPUTS
%
% OUTPUTS
% results_vec
% results_struct [structure] containing the following fields:
% posteriormean= posterior parameter mean
@ -44,16 +44,16 @@ function [results_vec, results_struct] = geweke_moments(draws,Dynareoptions)
% J.M. Bernardo, A.P. Dawid, and A.F.M. Smith (eds.) Proceedings of
% the Fourth Valencia International Meeting on Bayesian Statistics,
% pp. 169-194, Oxford University Press
% Geweke (1999): `Using simulation methods for Bayesian econometric models:
% Geweke (1999): `Using simulation methods for Bayesian econometric models:
% Inference, development and communication', Econometric Reviews, 18(1),
% 1-73
% -----------------------------------------------------------------
% written by: Johannes Pfeifer,
% based on code by James P. LeSage, who in turn
% drew on MATLAB programs written by Siddartha Chib
% written by: Johannes Pfeifer,
% based on code by James P. LeSage, who in turn
% drew on MATLAB programs written by Siddartha Chib
ndraw = size(draws,1);
n_groups=100;
taper_steps=Dynareoptions.convergence.geweke.taper_steps;
@ -66,7 +66,7 @@ window_means= zeros(n_groups,1);
window_uncentered_variances= zeros(n_groups,1);
for ig=1:n_groups
window_means(ig,1)=sum(draws((ig-1)*ns+1:ig*ns,1))/ns;
window_uncentered_variances(ig,1)=sum(draws((ig-1)*ns+1:ig*ns,1).^2)/ns;
window_uncentered_variances(ig,1)=sum(draws((ig-1)*ns+1:ig*ns,1).^2)/ns;
end %for ig
total_mean=mean(window_means);
total_variance=mean(window_uncentered_variances)-total_mean^2;
@ -106,4 +106,3 @@ for taper_index=1:length(taper_steps)
eval(['results_struct.nse_taper_',num2str(taper),'= NSE_taper;']);
eval(['results_struct.rne_taper_',num2str(taper),'= total_variance/(n_draws_used*NSE_taper^2);']);
end % end of for mm loop

142
matlab/convergence_diagnostics/mcmc_ifac.m
View File

@ -1,71 +1,71 @@
function Ifac = mcmc_ifac(X, Nc)
% function Ifac = mcmc_ifac(X, Nc)
% Compute inefficiency factor of a MCMC sample X based on a Parzen Window
%
% INPUTS
% X: time series
% Nc: # of lags
%
% OUTPUTS
% Ifac: inefficiency factor of MCMC sample
%
% SPECIAL REQUIREMENTS
% none
% ALGORITHM:
% Inefficiency factors are computed as
% \[
% Ifac = 1 + 2\sum\limits_{i=1}^{Nc} {\hat \rho(i)}
% \]
% where $\hat \rho(i)$ denotes the autocorrelation at lag i and the terms
% of the sum are truncated using a Parzen window.
%
% For inefficiency factors, see Section 6.1 of Paolo Giordani, Michael Pitt, and Robert Kohn (2011):
% "Bayesian Inference for Time Series State Space Models" in : John Geweke, Gary Koop,
% Herman van Dijk (editors): "The Oxford Handbook of Bayesian
% Econometrics", Oxford University Press
%
% The Parzen-Window is given by
% \[
% k(x) = \left\{ {\begin{array}{*{20}{c}}
% {1 - 6{x^2} + 6|x|^3} \text{ for } 0 \leqslant |x| \leqslant \frac{1}{2}} \\
% {2(1-|x|^3) \text{ for } \frac{1}{2} \leqslant |x| \leqslant 1} \\
% {0 \text{ otherwise}}
% \end{array}} \right.
% \]
% See Donald W.K Andrews (1991): "Heteroskedasticity and autocorrelation
% consistent covariance matrix estimation", Econometrica, 59(3), p. 817-858
% Copyright (C) 2015-16 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/>.
Nc = floor(min(Nc, length(X)/2));
if mod(Nc,2)
Nc=Nc-1;
end
AcorrXSIM = dyn_autocorr(X(:), Nc);
%
%Calculate the Parzen Weight
Parzen=zeros(Nc+1,1);
for i=1: Nc/2+1
Parzen(i)=1 - 6*(i/Nc)^2+ 6*(i/Nc)^3;
end
for i=(Nc/2)+1: Nc+1
Parzen(i)=2 * (1-(i/Nc))^3;
end
Parzen=Parzen';
Ifac= 1+2*sum(Parzen(:).* AcorrXSIM);
function Ifac = mcmc_ifac(X, Nc)
% function Ifac = mcmc_ifac(X, Nc)
% Compute inefficiency factor of a MCMC sample X based on a Parzen Window
%
% INPUTS
% X: time series
% Nc: # of lags
%
% OUTPUTS
% Ifac: inefficiency factor of MCMC sample
%
% SPECIAL REQUIREMENTS
% none
% ALGORITHM:
% Inefficiency factors are computed as
% \[
% Ifac = 1 + 2\sum\limits_{i=1}^{Nc} {\hat \rho(i)}
% \]
% where $\hat \rho(i)$ denotes the autocorrelation at lag i and the terms
% of the sum are truncated using a Parzen window.
%
% For inefficiency factors, see Section 6.1 of Paolo Giordani, Michael Pitt, and Robert Kohn (2011):
% "Bayesian Inference for Time Series State Space Models" in : John Geweke, Gary Koop,
% Herman van Dijk (editors): "The Oxford Handbook of Bayesian
% Econometrics", Oxford University Press
%
% The Parzen-Window is given by
% \[
% k(x) = \left\{ {\begin{array}{*{20}{c}}
% {1 - 6{x^2} + 6|x|^3} \text{ for } 0 \leqslant |x| \leqslant \frac{1}{2}} \\
% {2(1-|x|^3) \text{ for } \frac{1}{2} \leqslant |x| \leqslant 1} \\
% {0 \text{ otherwise}}
% \end{array}} \right.
% \]
% See Donald W.K Andrews (1991): "Heteroskedasticity and autocorrelation
% consistent covariance matrix estimation", Econometrica, 59(3), p. 817-858
% Copyright (C) 2015-16 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/>.
Nc = floor(min(Nc, length(X)/2));
if mod(Nc,2)
Nc=Nc-1;
end
AcorrXSIM = dyn_autocorr(X(:), Nc);
%
%Calculate the Parzen Weight
Parzen=zeros(Nc+1,1);
for i=1: Nc/2+1
Parzen(i)=1 - 6*(i/Nc)^2+ 6*(i/Nc)^3;
end
for i=(Nc/2)+1: Nc+1
Parzen(i)=2 * (1-(i/Nc))^3;
end
Parzen=Parzen';
Ifac= 1+2*sum(Parzen(:).* AcorrXSIM);

68
matlab/convergence_diagnostics/raftery_lewis.m
View File

@ -1,15 +1,15 @@
function [raftery_lewis] = raftery_lewis(runs,q,r,s)
% function raftery_lewis = raftery_lewis(runs,q,r,s)
% Computes the convergence diagnostics of Raftery and Lewis (1992), i.e. the
% number of draws needed in MCMC to estimate the posterior cdf of the q-quantile
% Computes the convergence diagnostics of Raftery and Lewis (1992), i.e. the
% number of draws needed in MCMC to estimate the posterior cdf of the q-quantile
% within an accuracy r with probability s
%
%
% Inputs:
% - draws [n_draws by n_var] double matrix of draws from the sampler
% - draws [n_draws by n_var] double matrix of draws from the sampler
% - q [scalar] quantile of the quantity of interest
% - r [scalar] level of desired precision
% - r [scalar] level of desired precision
% - s [scalar] probability associated with r
%
%
% Output:
% raftery_lewis [structure] containing the fields:
% - M_burn [n_draws by 1] number of draws required for burn-in
@ -21,7 +21,7 @@ function [raftery_lewis] = raftery_lewis(runs,q,r,s)
% iterations due to dependence in chain
% - N_min [scalar] # draws if the chain is white noise
% - N_total [n_draws by 1] nburn + nprec
%
%
% ---------------------------------------------------------------------
% NOTES: Example values of q, r, s:
@ -30,18 +30,18 @@ function [raftery_lewis] = raftery_lewis(runs,q,r,s)
%
% - The result is quite sensitive to r, being proportional to the
% inverse of r^2.
% - For epsilon (closeness of probabilities to equilibrium values),
% - For epsilon (closeness of probabilities to equilibrium values),
% Raftery/Lewis use 0.001 and argue that the results
% are quite robust to changes in this value
%
% ---------------------------------------------------------------------
% REFERENCES:
% REFERENCES:
% Raftery, Adrien E./Lewis, Steven (1992a): "How many iterations in the Gibbs sampler?"
% in: Bernardo/Berger/Dawid/Smith (eds.): Bayesian Statistics, Vol. 4, Clarendon Press: Oxford,
% pp. 763-773.
% Raftery, Adrien E./Lewis, Steven (1992b): "Comment: One long run with diagnostics:
% Implementation strategies for Markov chain Monte Carlo." Statistical Science,
% 7(4), pp. 493-497.
% in: Bernardo/Berger/Dawid/Smith (eds.): Bayesian Statistics, Vol. 4, Clarendon Press: Oxford,
% pp. 763-773.
% Raftery, Adrien E./Lewis, Steven (1992b): "Comment: One long run with diagnostics:
% Implementation strategies for Markov chain Monte Carlo." Statistical Science,
% 7(4), pp. 493-497.
%
% ----------------------------------------------------
@ -80,26 +80,26 @@ thinned_chain = zeros(n_runs,1);
Phi = norminv((s+1)/2); %note the missing ^{-1} at the Phi in equation top page 5, see RL (1995)
raftery_lewis.N_min = fix(Phi^2*(1-q)*q/r^2+1);
for nv = 1:n_vars % big loop over variables
for nv = 1:n_vars % big loop over variables
if q > 0 && q < 1
work = (runs(:,nv) <= quantile(runs(:,nv),q));
else
error('Quantile must be between 0 and 1');
error('Quantile must be between 0 and 1');
end
k_thin_current_var = 1;
bic = 1;
k_thin_current_var = 1;
bic = 1;
epss = 0.001;
% Find thinning factor for which first-order Markov Chain is preferred to second-order one
while(bic > 0)
thinned_chain=work(1:k_thin_current_var:n_runs,1);
thinned_chain=work(1:k_thin_current_var:n_runs,1);
[g2, bic] = first_vs_second_order_MC_test(thinned_chain);
k_thin_current_var = k_thin_current_var+1;
end
k_thin_current_var = k_thin_current_var-1; %undo last step
%compute transition probabilities
%compute transition probabilities
transition_matrix = zeros(2,2);
for i1 = 2:size(thinned_chain,1)
transition_matrix(thinned_chain(i1-1)+1,thinned_chain(i1)+1) = transition_matrix(thinned_chain(i1-1)+1,thinned_chain(i1)+1)+1;
@ -109,13 +109,13 @@ for nv = 1:n_vars % big loop over variables
kmind=k_thin_current_var;
[g2, bic]=independence_chain_test(thinned_chain);
while(bic > 0)
thinned_chain=work(1:kmind:n_runs,1);
thinned_chain=work(1:kmind:n_runs,1);
[g2, bic] = independence_chain_test(thinned_chain);
kmind = kmind+1;
end
m_star = log((alpha + beta)*epss/max(alpha,beta))/log(abs(1 - alpha - beta)); %equation bottom page 4
raftery_lewis.M_burn(nv) = fix((m_star+1)*k_thin_current_var);
n_star = (2 - (alpha + beta))*alpha*beta*(Phi^2)/((alpha + beta)^3 * r^2); %equation top page 5
@ -131,7 +131,7 @@ end
function [g2, bic] = first_vs_second_order_MC_test(d)
%conducts a test of first vs. second order Markov Chain via BIC criterion
n_obs=size(d,1);
g2 = 0;
g2 = 0;
tran=zeros(2,2,2);
for t_iter=3:n_obs % count state transitions
tran(d(t_iter-2,1)+1,d(t_iter-1,1)+1,d(t_iter,1)+1)=tran(d(t_iter-2,1)+1,d(t_iter-1,1)+1,d(t_iter,1)+1)+1;
@ -142,17 +142,17 @@ for ind_1 = 1:2
for ind_3 = 1:2
if tran(ind_1,ind_2,ind_3) ~= 0
fitted = (tran(ind_1,ind_2,1) + tran(ind_1,ind_2,2))*(tran(1,ind_2,ind_3) + tran(2,ind_2,ind_3))/...
(tran(1,ind_2,1) + tran(1,ind_2,2) + tran(2,ind_2,1) + tran(2,ind_2,2));
(tran(1,ind_2,1) + tran(1,ind_2,2) + tran(2,ind_2,1) + tran(2,ind_2,2));
focus = tran(ind_1,ind_2,ind_3);
g2 = g2 + log(focus/fitted)*focus;
end
end % end of for i3
end % end of for i2
end % end of for i1
g2 = g2*2;
bic = g2 - log(n_obs-2)*2;
g2 = g2*2;
bic = g2 - log(n_obs-2)*2;
end
end
function [g2, bic] = independence_chain_test(d)
@ -162,18 +162,18 @@ trans = zeros(2,2);
for ind_1 = 2:n_obs
trans(d(ind_1-1)+1,d(ind_1)+1)=trans(d(ind_1-1)+1,d(ind_1)+1)+1;
end
dcm1 = n_obs - 1;
dcm1 = n_obs - 1;
g2 = 0;
% Compute the log likelihood ratio statistic for second-order MC vs first-order MC. G2 statistic of Bishop, Fienberg and Holland (1975)
for ind_1 = 1:2
for ind_2 = 1:2
if trans(ind_1,ind_2) ~= 0
fitted = ((trans(ind_1,1) + trans(ind_1,2))*(trans(1,ind_2) + trans(2,ind_2)))/dcm1;
fitted = ((trans(ind_1,1) + trans(ind_1,2))*(trans(1,ind_2) + trans(2,ind_2)))/dcm1;
focus = trans(ind_1,ind_2);
g2 = g2 + log(focus/fitted)*focus;
end
end
end
g2 = g2*2;
g2 = g2*2;
bic = g2 - log(dcm1);
end

22
matlab/convert_dyn_45_to_44.m
View File

@ -40,7 +40,7 @@ if isfield(oo_,'PointForecast')
oo_.MeanForecast.(moment_names{moment_iter}).(var_names{var_iter})=...
[oo_.SmoothedVariables.(moment_names{moment_iter}).(var_names{var_iter})(:,end)*ones(M_.maximum_endo_lag,1) oo_.MeanForecast.(moment_names{moment_iter}).(var_names{var_iter})];
oo_.PointForecast.(moment_names{moment_iter}).(var_names{var_iter})=...
[oo_.SmoothedVariables.(moment_names{moment_iter}).(var_names{var_iter})(:,end)*ones(M_.maximum_endo_lag,1) oo_.PointForecast.(moment_names{moment_iter}).(var_names{var_iter})];
[oo_.SmoothedVariables.(moment_names{moment_iter}).(var_names{var_iter})(:,end)*ones(M_.maximum_endo_lag,1) oo_.PointForecast.(moment_names{moment_iter}).(var_names{var_iter})];
else
oo_.MeanForecast.(moment_names{moment_iter}).(var_names{var_iter})=...
[oo_.SmoothedVariables.(moment_names{moment_iter}).(var_names{var_iter})(end)*ones(M_.maximum_endo_lag,1); oo_.MeanForecast.(moment_names{moment_iter}).(var_names{var_iter})];
@ -106,7 +106,7 @@ if isfield(oo_,'UpdatedVariables')
for ii=1:length(names)
%make sure Bayesian fields are not affected
if ~strcmp(names{ii},'Mean') && ~strcmp(names{ii},'Median') && ~strcmp(names{ii},'deciles') ...
&& ~strcmp(names{ii},'Var') && ~strcmp(names{ii},'HPDinf') && ~strcmp(names{ii},'HPDsup')
&& ~strcmp(names{ii},'Var') && ~strcmp(names{ii},'HPDinf') && ~strcmp(names{ii},'HPDsup')
current_var_index=find(strmatch(names{ii},deblank(M_.endo_names),'exact'));
if options_.loglinear == 1 %logged steady state must be used
constant_current_variable=log(oo_.dr.ys(current_var_index));
@ -127,7 +127,7 @@ if isfield(oo_,'FilteredVariables')
for ii=1:length(names)
%make sure Bayesian fields are not affected
if ~strcmp(names{ii},'Mean') && ~strcmp(names{ii},'Median') && ~strcmp(names{ii},'deciles') ...
&& ~strcmp(names{ii},'Var') && ~strcmp(names{ii},'HPDinf') && ~strcmp(names{ii},'HPDsup')
&& ~strcmp(names{ii},'Var') && ~strcmp(names{ii},'HPDinf') && ~strcmp(names{ii},'HPDsup')
current_var_index=find(strmatch(names{ii},deblank(M_.endo_names),'exact'));
if options_.loglinear == 1 %logged steady state must be used
constant_current_variable=log(oo_.dr.ys(current_var_index));
@ -147,11 +147,11 @@ end
if ~isempty(options_.nk) && options_.nk ~= 0 && ~isempty(bayestopt_)
if ~((any(bayestopt_.pshape > 0) && options_.mh_replic) || (any(bayestopt_.pshape> 0) && options_.load_mh_file)) %no Bayesian estimation
positions_in_decision_order=oo_.dr.inv_order_var(bayestopt_.smoother_var_list(bayestopt_.smoother_saved_var_list));
if options_.loglinear == 1 %logged steady state must be used
constant_all_variables=log(oo_.dr.ys(bayestopt_.smoother_var_list(bayestopt_.smoother_saved_var_list)));
elseif options_.loglinear == 0 %unlogged steady state must be used
constant_all_variables=oo_.dr.ys(bayestopt_.smoother_var_list(bayestopt_.smoother_saved_var_list));
end
if options_.loglinear == 1 %logged steady state must be used
constant_all_variables=log(oo_.dr.ys(bayestopt_.smoother_var_list(bayestopt_.smoother_saved_var_list)));
elseif options_.loglinear == 0 %unlogged steady state must be used
constant_all_variables=oo_.dr.ys(bayestopt_.smoother_var_list(bayestopt_.smoother_saved_var_list));
end
if ~(options_.selected_variables_only && ~(options_.forecast > 0)) %happens only when selected_variables_only is not used
oo_.FilteredVariablesKStepAhead(:,positions_in_decision_order,:)=oo_.FilteredVariablesKStepAhead-constant_all_variables;
if ~isempty(PK) %get K-step ahead variances
@ -182,11 +182,11 @@ end
%Deal with OSR
if ~isempty(M_.osr.variable_weights)
evalin('base','optim_weights_=M_.osr.variable_weights')
evalin('base','optim_weights_=M_.osr.variable_weights')
end
if ~isempty(M_.osr.variable_indices)
evalin('base','obj_var_=M_.osr.variable_indices')
evalin('base','obj_var_=M_.osr.variable_indices')
end
if ~isempty(M_.osr.param_names)
evalin('base','osr_params_=char(M_.osr.param_names)')
evalin('base','osr_params_=char(M_.osr.param_names)')
end

2
matlab/convert_oo_.m
View File

@ -81,5 +81,5 @@ else
end
eval(['oo_ = convert_dyn_' strrep(new_from_ver, '.', '') '_to_' ...
strrep(to_ver, '.', '') '(M_, options_, oo_);']);
strrep(to_ver, '.', '') '(M_, options_, oo_);']);
end

2
matlab/correlation_mc_analysis.m
View File

@ -137,7 +137,7 @@ end
function oo_ = fill_output_structure(var1,var2,type,oo_,moment,lag,result)
switch moment
case {'Mean','Median','Variance','HPDinf','HPDsup'}
case {'Mean','Median','Variance','HPDinf','HPDsup'}
oo_.([type, 'TheoreticalMoments']).dsge.correlation.(moment).(var1).(var2)(lag,1) = result;
case {'deciles','density'}
oo_.([type, 'TheoreticalMoments']).dsge.correlation.(moment).(var1).(var2)(lag,1) = {result};

3
matlab/cosn.m
View File

@ -37,6 +37,3 @@ if rank(yhat)
else
co=0;
end

12
matlab/covariance_mc_analysis.m
View File

@ -1,7 +1,7 @@
function oo_ = covariance_mc_analysis(NumberOfSimulations,type,dname,fname,vartan,nvar,var1,var2,mh_conf_sig,oo_,options_)
% This function analyses the (posterior or prior) distribution of the
% endogenous variables' covariance matrix.
%
%
% INPUTS
% NumberOfSimulations [integer] scalar, number of simulations.
% type [string] 'prior' or 'posterior'
@ -14,7 +14,7 @@ function oo_ = covariance_mc_analysis(NumberOfSimulations,type,dname,fname,varta
% mh_conf_sig [double] 2 by 1 vector with upper
% and lower bound of HPD intervals
% oo_ [structure] Dynare structure where the results are saved.
% options_ [structure] Dynare options structure
% options_ [structure] Dynare options structure
%
% OUTPUTS
% oo_ [structure] Dynare structure where the results are saved.
@ -64,7 +64,7 @@ var1=deblank(var1);
var2=deblank(var2);
if isfield(oo_,[ TYPE 'TheoreticalMoments'])
temporary_structure = oo_.([TYPE, 'TheoreticalMoments']);
temporary_structure = oo_.([TYPE, 'TheoreticalMoments']);
if isfield(temporary_structure,'dsge')
temporary_structure = oo_.([TYPE, 'TheoreticalMoments']).dsge;
if isfield(temporary_structure,'covariance')
@ -104,17 +104,17 @@ for file = 1:length(ListOfFiles)
temp=Covariance_matrix(:,cov_pos)./(sqrt(Covariance_matrix(:,var_pos_1)).*sqrt(Covariance_matrix(:,var_pos_2)));
temp(Covariance_matrix(:,cov_pos)==0)=0; %filter out 0 correlations that would result in 0/0
tmp_corr_mat(i1:i2)=temp;
end
end
i1 = i2+1;
end
if options_.estimation.moments_posterior_density.indicator
[p_mean, p_median, p_var, hpd_interval, p_deciles, density] = ...
posterior_moments(tmp,1,mh_conf_sig);
posterior_moments(tmp,1,mh_conf_sig);
oo_.([TYPE, 'TheoreticalMoments']).dsge.covariance.density.(var1).(var2) = density;
else
[p_mean, p_median, p_var, hpd_interval, p_deciles] = ...
posterior_moments(tmp,0,mh_conf_sig);
posterior_moments(tmp,0,mh_conf_sig);
end
oo_.([TYPE, 'TheoreticalMoments']).dsge.covariance.Mean.(var1).(var2) = p_mean;
oo_.([TYPE, 'TheoreticalMoments']).dsge.covariance.Median.(var1).(var2) = p_median;

4
matlab/csolve.m
View File

@ -12,7 +12,7 @@ function [x,rc] = csolve(FUN,x,gradfun,crit,itmax,varargin)
% the equation is solved.
% itmax: the solver stops when this number of iterations is reached, with rc=4
% varargin: in this position the user can place any number of additional arguments, all
% of which are passed on to FUN and gradfun (when it is non-empty) as a list of
% of which are passed on to FUN and gradfun (when it is non-empty) as a list of
% arguments following x.
% rc: 0 means normal solution, 1 and 3 mean no solution despite extremely fine adjustments
% in step length (very likely a numerical problem, or a discontinuity). 4 means itmax
@ -60,7 +60,7 @@ if isempty(varargin)
f0=feval(FUN,x);
else
f0=feval(FUN,x,varargin{:});
end
end
af0=sum(abs(f0));
af00=af0;
itct=0;

32
matlab/cubature_with_gaussian_weight.m
View File

@ -29,8 +29,8 @@ function [nodes, weights] = cubature_with_gaussian_weight(d,n,method)
%! The routine returns nodes and associated weights to compute a multivariate integral of the form:
%!
%! \int_D f(x)*\exp(-<x,x>) dx
%!
%!
%!
%!
%! @end deftypefn
%@eod:
@ -52,7 +52,7 @@ function [nodes, weights] = cubature_with_gaussian_weight(d,n,method)
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
% AUTHOR(S) stephane DOT adjemian AT univ DASH lemans DOT fr
% Set default.
if nargin<3 || isempty(method)
method = 'Stroud';
@ -90,7 +90,7 @@ if strcmp(method,'Stroud') && isequal(n,5)
nodes = zeros(d,m);
weights = zeros(m,1);
% Set the weight for the first node (0)
weights(1) = A;
weights(1) = A;
skip = 1;
% Set the remaining nodes and associated weights.
nodes(:,skip+(1:d)) = r*eye(d);
@ -117,20 +117,20 @@ end
function v = e(n,i)
v = zeros(n,1);
v(i) = 1;
v = zeros(n,1);
v(i) = 1;
function m = ee(n,i,j)
m = zeros(n,4);
m(:,1) = e(n,i)+e(n,j);
m(:,2) = e(n,i)-e(n,j);
m(:,3) = -m(:,2);
m(:,4) = -m(:,1);
m = zeros(n,4);
m(:,1) = e(n,i)+e(n,j);
m(:,2) = e(n,i)-e(n,j);
m(:,3) = -m(:,2);
m(:,4) = -m(:,1);
%@test:1
%$ % Set problem
%$ d = 4;
%$
%$
%$ t = zeros(5,1);
%$
%$ % Call the tested routine
@ -289,7 +289,7 @@ function m = ee(n,i,j)
%@test:5
%$ % Set problem
%$ d = 5;
%$
%$
%$ t = zeros(6,1);
%$
%$ % Call the tested routine
@ -333,7 +333,7 @@ function m = ee(n,i,j)
%@test:6
%$ % Set problem
%$ d = 3;
%$
%$
%$ t = zeros(4,1);
%$
%$ % Call the tested routine

4
matlab/cycle_reduction.m
View File

@ -76,7 +76,7 @@ id0 = 1:n;
id2 = id0+n;
cont = 1;
while cont
while cont
tmp = ([A0; A2]/A1)*[A0 A2];
A1 = A1 - tmp(id0,id2) - tmp(id2,id0);
A0 = -tmp(id0,id0);
@ -97,7 +97,7 @@ while cont
info(2) = log(norm(A1,1));
end
return
end
end
it = it + 1;
end

15
matlab/datatomfile.m
View File

@ -3,16 +3,16 @@ function datatomfile (s, var_list, names)
% This command saves the simulation results in a text file. The name of each
% variable preceeds the corresponding results. This command must follow SIMUL
% or STOCH_SIMUL commands.
%
% INPUTS
%
% INPUTS
% - s: data file name
% - var_list: vector of selected endogenous variables
% - names: vector of strings (alternative names for the endogenous variables in the data file)
%
% OUTPUTS
% OUTPUTS
% none
%
% REMARKS
% REMARKS
% Only the first argument is mandatory. If only one input argument is
% provided, all the variables as defined in M_.endo_names will be saved in
% the generated m file.
@ -36,7 +36,7 @@ function datatomfile (s, var_list, names)
global M_ oo_
% Open the data file.
% Open the data file.
sm=[s,'.m'];
fid=fopen(sm,'w') ;
@ -65,8 +65,8 @@ for i=1:n
end
end
stack = dbstack;
fprintf(fid,'%% Dataset generated by %s.\n',stack(2).file);
fprintf(fid,['%% ' datestr(now,0) '\n']);
fprintf(fid,'%% Dataset generated by %s.\n',stack(2).file);
fprintf(fid,['%% ' datestr(now,0) '\n']);
% Save the selected data.
for i = 1:n
fprintf(fid,[strtrim(names(i,:)), ' = ['],'\n') ;
@ -78,4 +78,3 @@ end
% Close the data file.
fclose(fid) ;

2
matlab/dcompare.m
View File

@ -52,5 +52,3 @@ z = mean(mean(abs(x(j,i)-y(j,i)))) ;
disp (['The mean absolute difference between set ' s1(1,:) 'and set ' s1(2,:)]) ;
disp (['is : ' num2str(z)]) ;
return ;

2
matlab/delete_mh_history_files.m
View File

@ -1,5 +1,5 @@
function info = delete_mh_history_files(MetropolisFolder, ModelName)
% Copyright (C) 2013 Dynare Team
%
% This file is part of Dynare.

54
matlab/discretionary_policy_1.m
View File

@ -39,7 +39,7 @@ if isfield(M_,'orig_model')
M_.maximum_lag = orig_model.maximum_lag;
M_.maximum_endo_lag = orig_model.maximum_endo_lag;
else
M_.orig_model = M_;
M_.orig_model = M_;
end
beta = get_optimal_policy_discount_factor(M_.params,M_.param_names);
@ -53,10 +53,10 @@ if isfield(M_,'orig_model')
MaxLead = orig_model.maximum_lead;
MaxLag = orig_model.maximum_lag;
else
endo_names = M_.endo_names;
endo_nbr = M_.endo_nbr;
MaxLag=M_.maximum_lag;
MaxLead=M_.maximum_lead;
endo_names = M_.endo_names;
endo_nbr = M_.endo_nbr;
MaxLag=M_.maximum_lag;
MaxLead=M_.maximum_lead;
lead_lag_incidence = M_.lead_lag_incidence;
end
@ -64,7 +64,7 @@ end
if options_.steadystate_flag
% explicit steady state file
[junk,M_.params,info] = evaluate_steady_state_file(oo_.steady_state,[oo_.exo_steady_state; oo_.exo_det_steady_state],M_, ...
options_,0);
options_,0);
end
[U,Uy,W] = feval([M_.fname,'_objective_static'],zeros(endo_nbr,1),[], M_.params);
if any(any(Uy~=0))
@ -96,32 +96,32 @@ eq_nbr= size(jacobia_,1);
instr_nbr=endo_nbr-eq_nbr;
if instr_nbr==0
error('discretionary_policy:: There are no available instruments, because the model has as many equations as variables.')
error('discretionary_policy:: There are no available instruments, because the model has as many equations as variables.')
end
if size(Instruments,1)~= instr_nbr
error('discretionary_policy:: There are more declared instruments than omitted equations.')
end
error('discretionary_policy:: There are more declared instruments than omitted equations.')
end
instr_id=nan(instr_nbr,1);
for j=1:instr_nbr
vj=deblank(Instruments(j,:));
vj_id=strmatch(vj,endo_names,'exact');
if ~isempty(vj_id)
instr_id(j)=vj_id;
else
error([mfilename,':: instrument ',vj,' not found'])
end
vj=deblank(Instruments(j,:));
vj_id=strmatch(vj,endo_names,'exact');
if ~isempty(vj_id)
instr_id(j)=vj_id;
else
error([mfilename,':: instrument ',vj,' not found'])
end
end
Indices={'lag','0','lead'};
iter=1;
for j=1:numel(Indices)
eval(['A',Indices{j},'=zeros(eq_nbr,endo_nbr);'])
if strcmp(Indices{j},'0')||(strcmp(Indices{j},'lag') && MaxLag)||(strcmp(Indices{j},'lead') && MaxLead)
[junk,row,col]=find(lead_lag_incidence(iter,:));
eval(['A',Indices{j},'(:,row)=jacobia_(:,col);'])
iter=iter+1;
end
if strcmp(Indices{j},'0')||(strcmp(Indices{j},'lag') && MaxLag)||(strcmp(Indices{j},'lead') && MaxLead)
[junk,row,col]=find(lead_lag_incidence(iter,:));
eval(['A',Indices{j},'(:,row)=jacobia_(:,col);'])
iter=iter+1;
end
end
B=jacobia_(:,nnz(iyv)+1:end);
@ -131,17 +131,17 @@ solve_maxit = options_.dp.maxit;
discretion_tol = options_.discretionary_tol;
if ~isempty(Hold)
[H,G,info]=discretionary_policy_engine(Alag,A0,Alead,B,W,instr_id,beta,solve_maxit,discretion_tol,qz_criterium,Hold);
[H,G,info]=discretionary_policy_engine(Alag,A0,Alead,B,W,instr_id,beta,solve_maxit,discretion_tol,qz_criterium,Hold);
else
[H,G,info]=discretionary_policy_engine(Alag,A0,Alead,B,W,instr_id,beta,solve_maxit,discretion_tol,qz_criterium);
[H,G,info]=discretionary_policy_engine(Alag,A0,Alead,B,W,instr_id,beta,solve_maxit,discretion_tol,qz_criterium);
end
if info
dr=[];
return
else
Hold=H; %save previous solution
% Hold=[]; use this line if persistent command is not used.
Hold=H; %save previous solution
% Hold=[]; use this line if persistent command is not used.
end
% set the state
dr=oo_.dr;
@ -159,7 +159,7 @@ dr.ys=ys; % <--- dr.ys =zeros(NewEndo_nbr,1);
function ys=NondistortionarySteadyState(M_)
if exist([M_.fname,'_steadystate.m'],'file')
eval(['ys=',M_.fname,'_steadystate.m;'])
eval(['ys=',M_.fname,'_steadystate.m;'])
else
ys=zeros(M_.endo_nbr,1);
ys=zeros(M_.endo_nbr,1);
end

30
matlab/discretionary_policy_engine.m
View File

@ -1,23 +1,23 @@
function [H,G,retcode]=discretionary_policy_engine(AAlag,AA0,AAlead,BB,bigw,instr_id,beta,solve_maxit,discretion_tol,qz_criterium,H00,verbose)
% Solves the discretionary problem for a model of the form:
%
%
% Loss=E_0 sum_{t=0}^{\infty} beta^t [y_t'*W*y+x_t'*Q*x_t]
% subject to
% AAlag*yy_{t-1}+AA0*yy_t+AAlead*yy_{t+1}+BB*e=0
%
%
% with W the weight on the variables in vector y_t.
%
%
% The solution takes the form
% y_t=H*y_{t-1}+G*e_t
% where H=[H1;F1] and G=[H2;F2].
%
%
% We use the Dennis (2007, Macroeconomic Dynamics) algorithm and so we need
% to re-write the model in the form
% A0*y_t=A1*y_{t-1}+A2*y_{t+1}+A3*x_t+A4*x_{t+1}+A5*e_t, with W the
% weight on the y_t vector and Q the weight on the x_t vector of
% instruments.
%
%
% Inputs:
% AAlag [double] matrix of coefficients on lagged
% variables
@ -28,7 +28,7 @@ function [H,G,retcode]=discretionary_policy_engine(AAlag,AA0,AAlead,BB,bigw,inst
% BB [double] matrix of coefficients on
% shocks
% bigw [double] matrix of coefficients on variables in
% loss/objective function; stacks [W and Q]
% loss/objective function; stacks [W and Q]
% instr_id [double] location vector of the instruments in the yy_t vector.
% beta [scalar] planner discount factor
% solve_maxit [scalar] maximum number of iterations
@ -41,12 +41,12 @@ function [H,G,retcode]=discretionary_policy_engine(AAlag,AA0,AAlead,BB,bigw,inst
% H [double] (endo_nbr*endo_nbr) solution matrix for endogenous
% variables, stacks [H1 and H1]
% G [double] (endo_nbr*exo_nbr) solution matrix for shocks, stacks [H2 and F2]
%
%
% retcode [scalar] return code
%
% Algorithm:
% Dennis, Richard (2007): Optimal policy in rational expectations models: new solution algorithms,
% Macroeconomic Dynamics, 11, 31–55.
% Macroeconomic Dynamics, 11, 31–55.
% Copyright (C) 2007-2017 Dynare Team
%
@ -121,10 +121,10 @@ while 1
end
D=A0-A2*H1-A4*F1; %equation (20)
Dinv=inv(D);
A3DPD=A3'*Dinv'*P*Dinv;
A3DPD=A3'*Dinv'*P*Dinv;
F1=-(Q+A3DPD*A3)\(A3DPD*A1); %component of (26)
H1=Dinv*(A1+A3*F1); %component of (27)
[rcode,NQ]=CheckConvergence([H1;F1]-[H10;F10],iter,solve_maxit,discretion_tol);
if rcode
break
@ -174,7 +174,7 @@ else
H(instr_id,endo_augm_id)=F1;
G(endo_augm_id,:)=H2;
G(instr_id,:)=F2;
% Account for auxilliary variables
H(:,instr_id(aux))=H(:,end-(AuxiliaryVariables_nbr-1:-1:0));
H=H(1:endo_nbr,1:endo_nbr);
@ -253,7 +253,7 @@ end
function v = SylvesterHessenbergSchur(d,g,h)
%
% DSYLHS Solves a discrete time sylvester equation using the
% DSYLHS Solves a discrete time sylvester equation using the
% Hessenberg-Schur algorithm
%
% v = DSYLHS(g,d,h) computes the matrix v that satisfies the
@ -301,7 +301,7 @@ if i< n
i = i+1;
else
A = [w-g*h(i,i) (-g*h(i+1,i));...
-g*h(i,i+1) w-g*h(i+1,i+1)];
-g*h(i,i+1) w-g*h(i+1,i+1)];
C = [d(:,i); d(:,i+1)];
X = A\C;
v(:,i) = X(1:m,:);
@ -320,9 +320,9 @@ while i<n
i = i+1;
else
A = [w - g*h(i,i) (-g*h(i+1,i)); ...
-g*h(i,i+1) w - g*h(i+1,i+1)];
-g*h(i,i+1) w - g*h(i+1,i+1)];
C = [d(:,i) + temp*h(1:b,i); ...
d(:,i+1) + temp*h(1:b,i+1)];
d(:,i+1) + temp*h(1:b,i+1)];
X = A\C;
v(:,i) = X(1:m,:);
v(:,i+1) = X(m+1:2*m, :);

20
matlab/disp_dr.m
View File

@ -7,7 +7,7 @@ function disp_dr(dr,order,var_list)
% order [int]: order of approximation
% var_list [char array]: list of endogenous variables for which the
% decision rules should be printed
%
%
% Copyright (C) 2001-2017 Dynare Team
%
% This file is part of Dynare.
@ -80,7 +80,7 @@ var_name_width=max([max(size(deblank(M_.endo_names(k1(ivar),:)),2)),max(size(deb
%deal with covariances
if order > 1
var_name_width=max(2*(var_name_width+aux_var_additional_characters)+2,20); %account for covariances, separated by comma
var_name_width=max(2*(var_name_width+aux_var_additional_characters)+2,20); %account for covariances, separated by comma
else
var_name_width=max(var_name_width+aux_var_additional_characters,20);
end
@ -90,7 +90,7 @@ label_format = sprintf('%%-%ds',var_name_width);
%% start displayimg
disp('POLICY AND TRANSITION FUNCTIONS')
% variable names
str = char(32*ones(1,var_name_width));
str = char(32*ones(1,var_name_width));
for i=1:nvar
str = [str sprintf(header_label_format,deblank(M_.endo_names(k1(ivar(i)),:)))];
end
@ -232,7 +232,7 @@ for i = 1:length(M_.aux_vars)
if M_.aux_vars(i).endo_index == aux_index
switch M_.aux_vars(i).type
case 0
str = sprintf('%s(%d)',deblank(M_.endo_names(aux_index,:)),aux_lead_lag);
str = sprintf('%s(%d)',deblank(M_.endo_names(aux_index,:)),aux_lead_lag);
return
case 1
orig_name = deblank(M_.endo_names(M_.aux_vars(i).orig_index, :));
@ -256,10 +256,10 @@ error(sprintf('Could not find aux var: %s', M_.endo_names(aux_index, :)))
end
function [str,flag]=get_print_string(str,x,value_format_zero,value_format_float,flag,options_)
if abs(x) >= options_.dr_display_tol
flag = 1;
str = [str sprintf(value_format_float,x)];
else
str = [str sprintf(value_format_zero,0)];
end
if abs(x) >= options_.dr_display_tol
flag = 1;
str = [str sprintf(value_format_float,x)];
else
str = [str sprintf(value_format_zero,0)];
end
end

90
matlab/disp_identification.m
View File

@ -118,7 +118,7 @@ if any(idemodel.ino)
disp([' [',name{jx(1)},',',name{jy(1)},'] are PAIRWISE collinear (with tol = 1.e-10) !' ])
end
end
end
skipline()
for j=1:npar
@ -133,23 +133,23 @@ if any(idemodel.ino)
end
end
end
% if npar>(j+1),
% [ipair, jpair] = find(squeeze(idemodel.Pco(j,j+1:end,:))'>(1-1.e-10));
% else
% [ipair, jpair] = find(squeeze(idemodel.Pco(j,j+1:end,:))>(1-1.e-10));
% end
% if ~isempty(jpair),
% for jx=j+1:npar,
% ixp = find(jx==(jpair+j));
% if ~isempty(ixp)
% if SampleSize > 1,
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10) for ',num2str(length(ixp)/SampleSize*100),'% of MC runs!' ])
% else
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10)!' ])
% end
% end
% end
% end
% if npar>(j+1),
% [ipair, jpair] = find(squeeze(idemodel.Pco(j,j+1:end,:))'>(1-1.e-10));
% else
% [ipair, jpair] = find(squeeze(idemodel.Pco(j,j+1:end,:))>(1-1.e-10));
% end
% if ~isempty(jpair),
% for jx=j+1:npar,
% ixp = find(jx==(jpair+j));
% if ~isempty(ixp)
% if SampleSize > 1,
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10) for ',num2str(length(ixp)/SampleSize*100),'% of MC runs!' ])
% else
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10)!' ])
% end
% end
% end
% end
end
if ~any(idemodel.ino) && ~any(any(idemodel.ind0==0))
@ -167,10 +167,10 @@ if any(idemoments.ino)
end
% disp('WARNING !!!')
% disp(['The rank of J (moments) is deficient for ', num2str(length(find(idemoments.ino))/SampleSize*100),'% of MC runs!' ]),
% indno=[];
% for j=1:SampleSize, indno=[indno;idemoments.indno{j}]; end
% freqno = mean(indno)*100;
% ifreq=find(freqno);
% indno=[];
% for j=1:SampleSize, indno=[indno;idemoments.indno{j}]; end
% freqno = mean(indno)*100;
% ifreq=find(freqno);
% disp('MOMENT RANK FAILURE DUE TO COLLINEARITY OF PARAMETERS:');
skipline()
for j=1:npar
@ -199,7 +199,7 @@ if any(idemoments.ino)
disp([' [',name{jx(1)},',',name{jy(1)},'] are PAIRWISE collinear (with tol = 1.e-10) !' ])
end
end
end
skipline()
for j=1:npar
@ -214,25 +214,25 @@ if any(idemoments.ino)
end
end
end
% if npar>(j+1),
% [ipair, jpair] = find(squeeze(idemoments.Pco(j,j+1:end,:))'>(1-1.e-10));
% else
% [ipair, jpair] = find(squeeze(idemoments.Pco(j,j+1:end,:))>(1-1.e-10));
% end
% if ~isempty(jpair),
% for jx=j+1:npar,
% ixp = find(jx==(jpair+j));
% if ~isempty(ixp)
% if SampleSize > 1
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10) for ',num2str(length(ixp)/SampleSize*100),'% of MC runs!' ])
% else
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10) !' ])
% end
% end
% end
% end
% end
% end
% if npar>(j+1),
% [ipair, jpair] = find(squeeze(idemoments.Pco(j,j+1:end,:))'>(1-1.e-10));
% else
% [ipair, jpair] = find(squeeze(idemoments.Pco(j,j+1:end,:))>(1-1.e-10));
% end
% if ~isempty(jpair),
% for jx=j+1:npar,
% ixp = find(jx==(jpair+j));
% if ~isempty(ixp)
% if SampleSize > 1
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10) for ',num2str(length(ixp)/SampleSize*100),'% of MC runs!' ])
% else
% disp([' [',name{j},',',name{jx},'] are PAIRWISE collinear (with tol = 1.e-10) !' ])
% end
% end
% end
% end
% end
% end
end
if ~any(idemoments.ino) && ~any(any(idemoments.ind0==0))
skipline()
@ -255,12 +255,12 @@ end
% if advanced && (~options_.noprint),
% for j=1:length(kokP),
% dyntable([name{kokP(j)},' pairwise correlations in the model'],char(' ','min','mean','max'), ...
% char(name{jpM{j}}),[pminM{j}' pmeanM{j}' pmaxM{j}'],10,10,3);
% char(name{jpM{j}}),[pminM{j}' pmeanM{j}' pmaxM{j}'],10,10,3);
% end
%
%
% for j=1:length(kokPJ),
% dyntable([name{kokPJ(j)},' pairwise correlations in J moments'],char(' ','min','mean','max'), ...
% char(name{jpJ{j}}),[pminJ{j}' pmeanJ{j}' pmaxJ{j}'],10,10,3);
% char(name{jpJ{j}}),[pminJ{j}' pmeanJ{j}' pmaxJ{j}'],10,10,3);
% end
% end
% disp(' ')

3
matlab/disp_model_summary.m
View File

@ -4,7 +4,7 @@ function disp_model_summary(M,dr,options)
% displays the model summary
%
% INPUTS
% M [matlab structure] Definition of the model.
% M [matlab structure] Definition of the model.
% dr [matlab structure] Decision rules
% options [matlab structure] Options
%
@ -40,4 +40,3 @@ labels = deblank(M.exo_names);
headers = char('Variables',labels);
lh = size(labels,2)+2;
dyntable(options,my_title,headers,labels,M.Sigma_e,lh,10,6);

30
matlab/disp_moments.m
View File

@ -7,7 +7,7 @@ function oo_=disp_moments(y,var_list,M_,options_,oo_)
% M_ [structure] Dynare's model structure
% oo_ [structure] Dynare's results structure
% options_ [structure] Dynare's options structure
%
%
% OUTPUTS
% oo_ [structure] Dynare's results structure,
@ -57,18 +57,18 @@ s2 = mean(y.*y);
s = sqrt(s2);
oo_.mean = transpose(m);
oo_.var = y'*y/size(y,1);
oo_.skewness = (mean(y.^3)./s2.^1.5)';
oo_.skewness = (mean(y.^3)./s2.^1.5)';
oo_.kurtosis = (mean(y.^4)./(s2.*s2)-3)';
labels = deblank(M_.endo_names(ivar,:));
labels_TeX = deblank(M_.endo_names_tex(ivar,:));
if options_.nomoments == 0
z = [ m' s' s2' (mean(y.^3)./s2.^1.5)' (mean(y.^4)./(s2.*s2)-3)' ];
z = [ m' s' s2' (mean(y.^3)./s2.^1.5)' (mean(y.^4)./(s2.*s2)-3)' ];
title='MOMENTS OF SIMULATED VARIABLES';
title=add_filter_subtitle(title,options_);
headers=char('VARIABLE','MEAN','STD. DEV.','VARIANCE','SKEWNESS', ...
'KURTOSIS');
dyntable(options_,title,headers,labels,z,size(labels,2)+2,16,6);
@ -79,7 +79,7 @@ end
if options_.nocorr == 0
corr = (y'*y/size(y,1))./(s'*s);
if options_.contemporaneous_correlation
if options_.contemporaneous_correlation
oo_.contemporaneous_correlation = corr;
end
if options_.noprint == 0
@ -98,7 +98,7 @@ if options_.nocorr == 0
end
if options_.noprint == 0 && length(options_.conditional_variance_decomposition)
fprintf('\nSTOCH_SIMUL: conditional_variance_decomposition requires theoretical moments, i.e. periods=0.\n')
fprintf('\nSTOCH_SIMUL: conditional_variance_decomposition requires theoretical moments, i.e. periods=0.\n')
end
ar = options_.ar;
@ -119,7 +119,7 @@ if ar > 0
dyn_latex_table(M_,options_,title,'sim_autocorr_matrix',headers,labels_TeX,autocorr,size(labels_TeX,2)+2,8,4);
end
end
end
@ -143,7 +143,7 @@ if ~options_.nodecomposition
chol_S = chol(M_.Sigma_e(i_exo_var,i_exo_var)); %decompose rest
shock_mat=zeros(options_.periods,M_.exo_nbr); %initialize
shock_mat(:,i_exo_var)=oo_.exo_simul(:,i_exo_var)/chol_S; %invert construction of oo_.exo_simul from simult.m
for shock_iter=1:length(i_exo_var)
temp_shock_mat=zeros(size(shock_mat));
temp_shock_mat(:,i_exo_var(shock_iter))=shock_mat(:,i_exo_var(shock_iter));
@ -156,9 +156,9 @@ if ~options_.nodecomposition
if ~options_.noprint %options_.nomoments == 0
skipline()
title='VARIANCE DECOMPOSITION SIMULATING ONE SHOCK AT A TIME (in percent)';
title=add_filter_subtitle(title,options_);
headers = M_.exo_names;
headers(M_.exo_names_orig_ord,:) = headers;
headers = char(' ',headers);
@ -181,12 +181,12 @@ if ~options_.nodecomposition
end
end
warning(warning_old_state);
end
function y=get_filtered_time_series(y,m,options_)
if options_.hp_filter && ~options_.one_sided_hp_filter && ~options_.bandpass.indicator
[hptrend,y] = sample_hp_filter(y,options_.hp_filter);
elseif ~options_.hp_filter && options_.one_sided_hp_filter && ~options_.bandpass.indicator
@ -197,8 +197,8 @@ elseif ~options_.hp_filter && ~options_.one_sided_hp_filter && options_.bandpass
y=data_temp.data;
elseif ~options_.hp_filter && ~options_.one_sided_hp_filter && ~options_.bandpass.indicator
y = bsxfun(@minus, y, m);
else
else
error('disp_moments:: You cannot use more than one filter at the same time')
end
end

4
matlab/disp_steady_state.m
View File

@ -1,11 +1,11 @@
function disp_steady_state(M,oo)
% function disp_steady_state(M,oo)
% computes and prints the steady state calculations
%
%
% INPUTS
% M structure of parameters
% oo structure of results
%
%
% OUTPUTS
% none
%

24
matlab/disp_th_moments.m
View File

@ -76,7 +76,7 @@ if size(stationary_vars, 1) > 0
if M_.exo_nbr > 1 && ~nodecomposition
skipline()
if options_.order == 2
title='APPROXIMATED VARIANCE DECOMPOSITION (in percent)';
title='APPROXIMATED VARIANCE DECOMPOSITION (in percent)';
else
title='VARIANCE DECOMPOSITION (in percent)';
end
@ -86,7 +86,7 @@ if size(stationary_vars, 1) > 0
headers = char(' ',headers);
lh = size(deblank(M_.endo_names(ivar(stationary_vars),:)),2)+2;
dyntable(options_,title,headers,deblank(M_.endo_names(ivar(stationary_vars), ...
:)),100* ...
:)),100* ...
oo_.gamma_y{options_.ar+2}(stationary_vars,:),lh,8,2);
if options_.TeX
headers=M_.exo_names_tex;
@ -97,7 +97,7 @@ if size(stationary_vars, 1) > 0
end
end
end
conditional_variance_steps = options_.conditional_variance_decomposition;
if length(conditional_variance_steps)
StateSpaceModel.number_of_state_equations = M_.endo_nbr;
@ -107,10 +107,10 @@ if size(stationary_vars, 1) > 0
StateSpaceModel.state_innovations_covariance_matrix = M_.Sigma_e;
StateSpaceModel.order_var = dr.order_var;
oo_.conditional_variance_decomposition = conditional_variance_decomposition(StateSpaceModel,conditional_variance_steps,ivar);
if options_.noprint == 0
display_conditional_variance_decomposition(oo_.conditional_variance_decomposition,conditional_variance_steps,...
ivar,M_,options_);
ivar,M_,options_);
end
end
end
@ -124,13 +124,13 @@ end
if options_.nocorr == 0 && size(stationary_vars, 1) > 0
corr = oo_.gamma_y{1}(i1,i1)./(sd(i1)*sd(i1)');
if options_.contemporaneous_correlation
if options_.contemporaneous_correlation
oo_.contemporaneous_correlation = corr;
end
if ~options_.noprint
skipline()
if options_.order == 2
title='APPROXIMATED MATRIX OF CORRELATIONS';
title='APPROXIMATED MATRIX OF CORRELATIONS';
else
title='MATRIX OF CORRELATIONS';
end
@ -154,22 +154,22 @@ if options_.ar > 0 && size(stationary_vars, 1) > 0
z(:,i) = diag(oo_.gamma_y{i+1}(i1,i1));
end
if ~options_.noprint
skipline()
skipline()
if options_.order == 2
title='APPROXIMATED COEFFICIENTS OF AUTOCORRELATION';
title='APPROXIMATED COEFFICIENTS OF AUTOCORRELATION';
else
title='COEFFICIENTS OF AUTOCORRELATION';
end
title=add_filter_subtitle(title,options_);
labels = deblank(M_.endo_names(ivar(i1),:));
labels = deblank(M_.endo_names(ivar(i1),:));
headers = char('Order ',int2str([1:options_.ar]'));
lh = size(labels,2)+2;
dyntable(options_,title,headers,labels,z,lh,8,4);
if options_.TeX
labels = deblank(M_.endo_names_tex(ivar(i1),:));
labels = deblank(M_.endo_names_tex(ivar(i1),:));
headers=char('Order ',int2str([1:options_.ar]'));
lh = size(labels,2)+2;
dyn_latex_table(M_,options_,title,'th_autocorr_matrix',headers,labels,z,lh,8,4);
end
end
end
end

18
matlab/display_conditional_variance_decomposition.m
View File

@ -1,8 +1,8 @@
function display_conditional_variance_decomposition(conditional_decomposition_array,Steps,SubsetOfVariables,M_,options_)
% This function displays the conditional variance decomposition of a given state space model
% for a subset of endogenous variables.
%
% INPUTS
%
% INPUTS
% conditional_decomposition_array [matrix] Output matrix from compute_conditional_variance_decomposition
% Steps [integer] 1*h vector of dates.
% SubsetOfVariables [integer] 1*q vector of indices.
@ -10,7 +10,7 @@ function display_conditional_variance_decomposition(conditional_decomposition_ar
% Model description
% options_ [structure] Dynare structure containing the
% options
% OUTPUTS
% OUTPUTS
% none
%
% Copyright (C) 2010-2016 Dynare Team
@ -56,10 +56,10 @@ for i=1:length(Steps)
dyntable(options_,'',headers,...
deblank(M_.endo_names(SubsetOfVariables,:)),...
vardec_i,lh,8,2);
if options_.TeX
labels_TeX = deblank(M_.endo_names_tex(SubsetOfVariables,:));
headers_TeX=char('',deblank(M_.exo_names_tex));
lh = size(labels_TeX,2)+2;
dyn_latex_table(M_,options_,[title,'; Period ' int2str(Steps(i))],['th_var_decomp_cond_h',int2str(Steps(i))],headers_TeX,labels_TeX,vardec_i,lh,8,2);
end
if options_.TeX
labels_TeX = deblank(M_.endo_names_tex(SubsetOfVariables,:));
headers_TeX=char('',deblank(M_.exo_names_tex));
lh = size(labels_TeX,2)+2;
dyn_latex_table(M_,options_,[title,'; Period ' int2str(Steps(i))],['th_var_decomp_cond_h',int2str(Steps(i))],headers_TeX,labels_TeX,vardec_i,lh,8,2);
end
end

170
matlab/display_estimation_results_table.m
View File

@ -1,20 +1,20 @@
function oo_=display_estimation_results_table(xparam1,stdh,M_,options_,estim_params_,bayestopt_,oo_,pnames,table_title,field_name)
%function oo_=display_results_table(xparam1,stdh,M_,estim_params_,bayestopt_,oo_,pnames,table_title,field_name)
% Display estimation results on screen and write them to TeX-file
%
% INPUTS
%
% INPUTS
% o xparam1 [double] (p*1) vector of estimate parameters.
% o stdh [double] (p*1) vector of estimate parameters.
% o M_ Matlab's structure describing the Model (initialized by dynare, see @ref{M_}).
% o M_ Matlab's structure describing the Model (initialized by dynare, see @ref{M_}).
% o estim_params_ Matlab's structure describing the estimated_parameters (initialized by dynare, see @ref{estim_params_}).
% o options_ Matlab's structure describing the options (initialized by dynare, see @ref{options_}).
% o bayestopt_ Matlab's structure describing the priors (initialized by dynare, see @ref{bayesopt_}).
% o oo_ Matlab's structure gathering the results (initialized by dynare, see @ref{oo_}).
% o pnames [string] Character Array storing the names for prior distributions
% o table_title [string] Title of the Table
% o pnames [string] Character Array storing the names for prior distributions
% o table_title [string] Title of the Table
% o field_name [string] String storing the name of the fields for oo_ where the parameters are stored
%
% OUTPUTS
%
% OUTPUTS
% o oo_ Matlab's structure gathering the results
%
% SPECIAL REQUIREMENTS
@ -64,13 +64,13 @@ if np
name = bayestopt_.name{ip};
if strcmp(field_name,'posterior')
fprintf('%-*s %7.3f %8.4f %7.4f %4s %6.4f \n', ...
header_width,name, ...
bayestopt_.p1(ip),xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip));
header_width,name, ...
bayestopt_.p1(ip),xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip));
else
fprintf('%-*s %8.4f %7.4f %7.4f \n', ...
header_width,name,xparam1(ip),stdh(ip),tstath(ip));
header_width,name,xparam1(ip),stdh(ip),tstath(ip));
end
eval(['oo_.' field_name '_mode.parameters.' name ' = xparam1(ip);']);
eval(['oo_.' field_name '_std_at_mode.parameters.' name ' = stdh(ip);']);
@ -87,9 +87,9 @@ if nvx
name = deblank(M_.exo_names(k,:));
if strcmp(field_name,'posterior')
fprintf('%-*s %7.3f %8.4f %7.4f %4s %6.4f \n', ...
header_width,name,bayestopt_.p1(ip),xparam1(ip), ...
stdh(ip),pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip));
header_width,name,bayestopt_.p1(ip),xparam1(ip), ...
stdh(ip),pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip));
else
fprintf('%-*s %8.4f %7.4f %7.4f \n',header_width,name,xparam1(ip),stdh(ip),tstath(ip));
end
@ -99,28 +99,28 @@ if nvx
ip = ip+1;
end
skipline()
end
if nvn
end
if nvn
disp('standard deviation of measurement errors')
disp(tit1)
ip = nvx+1;
for i=1:nvn
name = options_.varobs{estim_params_.nvn_observable_correspondence(i,1)};
if strcmp(field_name,'posterior')
if strcmp(field_name,'posterior')
fprintf('%-*s %7.3f %8.4f %7.4f %4s %6.4f \n', ...
header_width,name,bayestopt_.p1(ip), ...
xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip));
header_width,name,bayestopt_.p1(ip), ...
xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), ...
bayestopt_.p2(ip));
else
fprintf('%-*s %8.4f %7.4f %7.4f \n',header_width,name,xparam1(ip),stdh(ip),tstath(ip))
fprintf('%-*s %8.4f %7.4f %7.4f \n',header_width,name,xparam1(ip),stdh(ip),tstath(ip))
end
eval(['oo_.' field_name '_mode.measurement_errors_std.' name ' = xparam1(ip);']);
eval(['oo_.' field_name '_std_at_mode.measurement_errors_std.' name ' = stdh(ip);']);
ip = ip+1;
end
skipline()
end
end
if ncx
disp('correlation of shocks')
@ -131,12 +131,12 @@ if ncx
k2 = estim_params_.corrx(i,2);
name = [deblank(M_.exo_names(k1,:)) ',' deblank(M_.exo_names(k2,:))];
NAME = [deblank(M_.exo_names(k1,:)) '_' deblank(M_.exo_names(k2,:))];
if strcmp(field_name,'posterior')
if strcmp(field_name,'posterior')
fprintf('%-*s %7.3f %8.4f %7.4f %4s %6.4f \n', ...
header_width,name,bayestopt_.p1(ip),xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), bayestopt_.p2(ip));
header_width,name,bayestopt_.p1(ip),xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), bayestopt_.p2(ip));
else
fprintf('%-*s %8.4f %7.4f %7.4f \n', header_width,name,xparam1(ip),stdh(ip),tstath(ip));
fprintf('%-*s %8.4f %7.4f %7.4f \n', header_width,name,xparam1(ip),stdh(ip),tstath(ip));
end
M_.Sigma_e(k1,k2) = xparam1(ip)*sqrt(M_.Sigma_e(k1,k1)*M_.Sigma_e(k2,k2));
M_.Sigma_e(k2,k1) = M_.Sigma_e(k1,k2);
@ -156,12 +156,12 @@ if ncn
k2 = estim_params_.corrn(i,2);
name = [deblank(M_.endo_names(k1,:)) ',' deblank(M_.endo_names(k2,:))];
NAME = [deblank(M_.endo_names(k1,:)) '_' deblank(M_.endo_names(k2,:))];
if strcmp(field_name,'posterior')
if strcmp(field_name,'posterior')
fprintf('%-*s %7.3f %8.4f %7.4f %4s %6.4f \n', ...
header_width,name,bayestopt_.p1(ip),xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), bayestopt_.p2(ip));
header_width,name,bayestopt_.p1(ip),xparam1(ip),stdh(ip), ...
pnames(bayestopt_.pshape(ip)+1,:), bayestopt_.p2(ip));
else
fprintf('%-*s %8.4f %7.4f %7.4f \n',header_width,name,xparam1(ip),stdh(ip),tstath(ip));
fprintf('%-*s %8.4f %7.4f %7.4f \n',header_width,name,xparam1(ip),stdh(ip),tstath(ip));
end
eval(['oo_.' field_name '_mode.measurement_errors_corr.' NAME ' = xparam1(ip);']);
eval(['oo_.' field_name '_std_at_mode.measurement_errors_corr.' NAME ' = stdh(ip);']);
@ -268,17 +268,17 @@ if any(bayestopt_.pshape > 0) && options_.TeX %% Bayesian estimation (posterior
TeXEnd(fidTeX)
end
elseif all(bayestopt_.pshape == 0) && options_.TeX %% MLE and GMM Latex output
if np
if np
filename = [OutputDirectoryName '/' M_.fname '_' LaTeXtitle '_Mode_1.tex'];
fidTeX = fopen(filename,'w');
TeXBegin_ML(fidTeX,1,'parameters',table_title,LaTeXtitle)
TeXBegin_ML(fidTeX,1,'parameters',table_title,LaTeXtitle)
ip = nvx+nvn+ncx+ncn+1;
for i=1:np
fprintf(fidTeX,'$%s$ & %8.4f & %7.4f & %7.4f\\\\ \n',...
M_.param_names_tex(estim_params_.param_vals(i,1),:),...
xparam1(ip),...
stdh(ip),...
tstath(ip));
tstath(ip));
ip = ip + 1;
end
TeXEnd(fidTeX)
@ -286,7 +286,7 @@ elseif all(bayestopt_.pshape == 0) && options_.TeX %% MLE and GMM Latex output
if nvx
filename = [OutputDirectoryName '/' M_.fname '_' LaTeXtitle '_Mode_2.tex'];
fidTeX = fopen(filename,'w');
TeXBegin_ML(fidTeX,2,'standard deviation of structural shocks',table_title,LaTeXtitle)
TeXBegin_ML(fidTeX,2,'standard deviation of structural shocks',table_title,LaTeXtitle)
ip = 1;
for i=1:nvx
k = estim_params_.var_exo(i,1);
@ -302,11 +302,11 @@ elseif all(bayestopt_.pshape == 0) && options_.TeX %% MLE and GMM Latex output
if nvn
filename = [OutputDirectoryName '/' M_.fname '_' LaTeXtitle '_Mode_3.tex'];
fidTeX = fopen(filename,'w');
TeXBegin_ML(fidTeX,3,'standard deviation of measurement errors',table_title,LaTeXtitle)
TeXBegin_ML(fidTeX,3,'standard deviation of measurement errors',table_title,LaTeXtitle)
ip = nvx+1;
for i=1:nvn
idx = strmatch(options_.varobs{estim_params_.nvn_observable_correspondence(i,1)},M_.endo_names);
fprintf(fidTeX,'$%s$ & %8.4f & %7.4f & %7.4f \\\\ \n',...
idx = strmatch(options_.varobs{estim_params_.nvn_observable_correspondence(i,1)},M_.endo_names);
fprintf(fidTeX,'$%s$ & %8.4f & %7.4f & %7.4f \\\\ \n',...
deblank(M_.endo_names_tex(idx,:)), ...
xparam1(ip),...
stdh(ip),...
@ -318,7 +318,7 @@ elseif all(bayestopt_.pshape == 0) && options_.TeX %% MLE and GMM Latex output
if ncx
filename = [OutputDirectoryName '/' M_.fname '_' LaTeXtitle '_Mode_4.tex'];
fidTeX = fopen(filename,'w');
TeXBegin_ML(fidTeX,4,'correlation of structural shocks',table_title,LaTeXtitle)
TeXBegin_ML(fidTeX,4,'correlation of structural shocks',table_title,LaTeXtitle)
ip = nvx+nvn+1;
for i=1:ncx
k1 = estim_params_.corrx(i,1);
@ -335,7 +335,7 @@ elseif all(bayestopt_.pshape == 0) && options_.TeX %% MLE and GMM Latex output
if ncn
filename = [OutputDirectoryName '/' M_.fname '_' LaTeXtitle '_Mode_5.tex'];
fidTeX = fopen(filename,'w');
TeXBegin_ML(fidTeX,5,'correlation of measurement errors',table_title,LaTeXtitle)
TeXBegin_ML(fidTeX,5,'correlation of measurement errors',table_title,LaTeXtitle)
ip = nvx+nvn+ncx+1;
for i=1:ncn
k1 = estim_params_.corrn(i,1);
@ -347,7 +347,7 @@ elseif all(bayestopt_.pshape == 0) && options_.TeX %% MLE and GMM Latex output
tstath(ip));
ip = ip+1;
end
TeXEnd(fidTeX)
TeXEnd(fidTeX)
end
end
@ -356,51 +356,51 @@ end
%% subfunctions:
%
function TeXBegin_Bayesian(fid,fnum,title)
fprintf(fid,'%% TeX-table generated by dynare_estimation (Dynare).\n');
fprintf(fid,['%% RESULTS FROM POSTERIOR MAXIMIZATION (' title ')\n']);
fprintf(fid,['%% ' datestr(now,0)]);
fprintf(fid,' \n');
fprintf(fid,' \n');
fprintf(fid,'\\begin{center}\n');
fprintf(fid,'\\begin{longtable}{llcccc} \n');
fprintf(fid,['\\caption{Results from posterior maximization (' title ')}\\\\\n ']);
fprintf(fid,['\\label{Table:Posterior:' int2str(fnum) '}\\\\\n']);
fprintf(fid,'\\toprule \n');
fprintf(fid,' & \\multicolumn{3}{c}{Prior} & \\multicolumn{2}{c}{Posterior} \\\\\n');
fprintf(fid,' \\cmidrule(r{.75em}){2-4} \\cmidrule(r{.75em}){5-6}\n');
fprintf(fid,' & Dist. & Mean & Stdev & Mode & Stdev \\\\ \n');
fprintf(fid,'\\midrule \\endfirsthead \n');
fprintf(fid,'\\caption{(continued)}\\\\\n ');
fprintf(fid,'\\bottomrule \n');
fprintf(fid,' & \\multicolumn{3}{c}{Prior} & \\multicolumn{2}{c}{Posterior} \\\\\n');
fprintf(fid,' \\cmidrule(r{.75em}){2-4} \\cmidrule(r{.75em}){5-6}\n');
fprintf(fid,' & Dist. & Mean & Stdev & Mode & Stdev \\\\ \n');
fprintf(fid,'\\midrule \\endhead \n');
fprintf(fid,'\\bottomrule \\multicolumn{6}{r}{(Continued on next page)}\\endfoot \n');
fprintf(fid,'\\bottomrule\\endlastfoot \n');
fprintf(fid,'%% TeX-table generated by dynare_estimation (Dynare).\n');
fprintf(fid,['%% RESULTS FROM POSTERIOR MAXIMIZATION (' title ')\n']);
fprintf(fid,['%% ' datestr(now,0)]);
fprintf(fid,' \n');
fprintf(fid,' \n');
fprintf(fid,'\\begin{center}\n');
fprintf(fid,'\\begin{longtable}{llcccc} \n');
fprintf(fid,['\\caption{Results from posterior maximization (' title ')}\\\\\n ']);
fprintf(fid,['\\label{Table:Posterior:' int2str(fnum) '}\\\\\n']);
fprintf(fid,'\\toprule \n');
fprintf(fid,' & \\multicolumn{3}{c}{Prior} & \\multicolumn{2}{c}{Posterior} \\\\\n');
fprintf(fid,' \\cmidrule(r{.75em}){2-4} \\cmidrule(r{.75em}){5-6}\n');
fprintf(fid,' & Dist. & Mean & Stdev & Mode & Stdev \\\\ \n');
fprintf(fid,'\\midrule \\endfirsthead \n');
fprintf(fid,'\\caption{(continued)}\\\\\n ');
fprintf(fid,'\\bottomrule \n');
fprintf(fid,' & \\multicolumn{3}{c}{Prior} & \\multicolumn{2}{c}{Posterior} \\\\\n');
fprintf(fid,' \\cmidrule(r{.75em}){2-4} \\cmidrule(r{.75em}){5-6}\n');
fprintf(fid,' & Dist. & Mean & Stdev & Mode & Stdev \\\\ \n');
fprintf(fid,'\\midrule \\endhead \n');
fprintf(fid,'\\bottomrule \\multicolumn{6}{r}{(Continued on next page)}\\endfoot \n');
fprintf(fid,'\\bottomrule\\endlastfoot \n');
function TeXBegin_ML(fid,fnum,title,table_title,LaTeXtitle)
fprintf(fid,'%% TeX-table generated by dynare_estimation (Dynare).\n');
fprintf(fid,['%% RESULTS FROM ' table_title ' MAXIMIZATION (' title ')\n']);
fprintf(fid,['%% ' datestr(now,0)]);
fprintf(fid,' \n');
fprintf(fid,' \n');
fprintf(fid,'\\begin{center}\n');
fprintf(fid,'\\begin{longtable}{llcc} \n');
fprintf(fid,['\\caption{Results from ' table_title ' maximization (' title ')}\\\\\n ']);
fprintf(fid,['\\label{Table:' LaTeXtitle ':' int2str(fnum) '}\\\\\n']);
fprintf(fid,'\\toprule \n');
fprintf(fid,' & Mode & s.d. & t-stat\\\\ \n');
fprintf(fid,'\\midrule \\endfirsthead \n');
fprintf(fid,'\\caption{(continued)}\\\\\n ');
fprintf(fid,'\\toprule \n');
fprintf(fid,' & Mode & s.d. & t-stat\\\\ \n');
fprintf(fid,'\\midrule \\endhead \n');
fprintf(fid,'\\bottomrule \\multicolumn{4}{r}{(Continued on next page)} \\endfoot \n');
fprintf(fid,'\\bottomrule \\endlastfoot \n');
function TeXBegin_ML(fid,fnum,title,table_title,LaTeXtitle)
fprintf(fid,'%% TeX-table generated by dynare_estimation (Dynare).\n');
fprintf(fid,['%% RESULTS FROM ' table_title ' MAXIMIZATION (' title ')\n']);
fprintf(fid,['%% ' datestr(now,0)]);
fprintf(fid,' \n');
fprintf(fid,' \n');
fprintf(fid,'\\begin{center}\n');
fprintf(fid,'\\begin{longtable}{llcc} \n');
fprintf(fid,['\\caption{Results from ' table_title ' maximization (' title ')}\\\\\n ']);
fprintf(fid,['\\label{Table:' LaTeXtitle ':' int2str(fnum) '}\\\\\n']);
fprintf(fid,'\\toprule \n');
fprintf(fid,' & Mode & s.d. & t-stat\\\\ \n');
fprintf(fid,'\\midrule \\endfirsthead \n');
fprintf(fid,'\\caption{(continued)}\\\\\n ');
fprintf(fid,'\\toprule \n');
fprintf(fid,' & Mode & s.d. & t-stat\\\\ \n');
fprintf(fid,'\\midrule \\endhead \n');
fprintf(fid,'\\bottomrule \\multicolumn{4}{r}{(Continued on next page)} \\endfoot \n');
fprintf(fid,'\\bottomrule \\endlastfoot \n');
function TeXEnd(fid)
fprintf(fid,'\\end{longtable}\n ');
fprintf(fid,'\\end{longtable}\n ');
fprintf(fid,'\\end{center}\n');
fprintf(fid,'%% End of TeX file.\n');
fclose(fid);

74
matlab/display_problematic_vars_Jacobian.m
View File

@ -1,20 +1,20 @@
function []=display_problematic_vars_Jacobian(problemrow,problemcol,M_,x,type,caller_string)
% []=display_problematic_vars_Jacobian(problemrow,problemcol,M_,ys,caller_string)
% print the equation numbers and variables associated with problematic entries
% of the Jacobian
% print the equation numbers and variables associated with problematic entries
% of the Jacobian
%
% INPUTS
% problemrow [vector] rows associated with problematic entries
% problemcol [vector] columns associated with problematic entries
% M_ [matlab structure] Definition of the model.
% M_ [matlab structure] Definition of the model.
% x [vector] point at which the Jacobian was evaluated
% type [string] 'static' or 'dynamic' depending on the type of
% Jacobian
% caller_string [string] contains name of calling function for printing
%
% caller_string [string] contains name of calling function for printing
%
% OUTPUTS
% none.
%
%
% Copyright (C) 2014-16 Dynare Team
%
@ -62,25 +62,25 @@ if strcmp(type,'dynamic')
fprintf('Derivative of Equation %d with respect to %s Variable %s (initial value of %s: %g) \n',eq_nbr,type_string,deblank(M_.endo_names(var_index,:)),deblank(M_.endo_names(var_index,:)),x(var_index))
end
elseif problemcol(ii)<=max(max(M_.lead_lag_incidence)) && var_index>M_.orig_endo_nbr %auxiliary vars
if M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).type ==6 %Ramsey Lagrange Multiplier
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
fprintf('Derivative of Auxiliary Equation %d with respect to %s of Langrange multiplier of equation %s (initial value: %g) \n',eq_nbr,type_string,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
fprintf('Derivative of Equation %d with respect to %s of Langrange multiplier of equation %s (initial value: %g) \n',eq_nbr,type_string,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
end
if M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).type ==6 %Ramsey Lagrange Multiplier
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
fprintf('Derivative of Auxiliary Equation %d with respect to %s of Langrange multiplier of equation %s (initial value: %g) \n',eq_nbr,type_string,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
else
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
orig_var_index=M_.aux_vars(1,var_index-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Auxiliary Equation %d with respect to %s Variable %s (initial value of %s: %g) \n',eq_nbr,type_string,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(orig_var_index))
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
orig_var_index=M_.aux_vars(1,var_index-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Equation %d with respect to %s Variable %s (initial value of %s: %g) \n',eq_nbr,type_string,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(orig_var_index))
end
eq_nbr=problemrow(ii)-aux_eq_nbr;
fprintf('Derivative of Equation %d with respect to %s of Langrange multiplier of equation %s (initial value: %g) \n',eq_nbr,type_string,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
end
else
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
orig_var_index=M_.aux_vars(1,var_index-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Auxiliary Equation %d with respect to %s Variable %s (initial value of %s: %g) \n',eq_nbr,type_string,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(orig_var_index))
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
orig_var_index=M_.aux_vars(1,var_index-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Equation %d with respect to %s Variable %s (initial value of %s: %g) \n',eq_nbr,type_string,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(orig_var_index))
end
end
elseif problemcol(ii)>max(max(M_.lead_lag_incidence)) && var_index<=M_.exo_nbr
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
@ -88,7 +88,7 @@ if strcmp(type,'dynamic')
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
fprintf('Derivative of Equation %d with respect to %s shock %s \n',eq_nbr,type_string,deblank(M_.exo_names(var_index,:)));
end
end
else
error('display_problematic_vars_Jacobian:: The error should not happen. Please contact the developers')
end
@ -108,24 +108,24 @@ elseif strcmp(type,'static')
fprintf('Derivative of Equation %d with respect to Variable %s (initial value of %s: %g) \n',eq_nbr,deblank(M_.endo_names(problemcol(ii),:)),deblank(M_.endo_names(problemcol(ii),:)),x(problemcol(ii)))
end
else %auxiliary vars
if M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).type ==6 %Ramsey Lagrange Multiplier
if M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).type ==6 %Ramsey Lagrange Multiplier
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
fprintf('Derivative of Auxiliary Equation %d with respect to Lagrange multiplier of equation %d (initial value: %g) \n',eq_nbr,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
fprintf('Derivative of Auxiliary Equation %d with respect to Lagrange multiplier of equation %d (initial value: %g) \n',eq_nbr,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
fprintf('Derivative of Equation %d with respect to Lagrange multiplier of equation %d (initial value: %g) \n',eq_nbr,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
fprintf('Derivative of Equation %d with respect to Lagrange multiplier of equation %d (initial value: %g) \n',eq_nbr,M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).eq_nbr,x(problemcol(ii)))
end
else
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
orig_var_index=M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Auxiliary Equation %d with respect to Variable %s (initial value of %s: %g) \n',eq_nbr,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(problemcol(ii)))
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
orig_var_index=M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Equation %d with respect to Variable %s (initial value of %s: %g) \n',eq_nbr,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(problemcol(ii)))
end
if problemrow(ii)<=aux_eq_nbr
eq_nbr=problemrow(ii);
orig_var_index=M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Auxiliary Equation %d with respect to Variable %s (initial value of %s: %g) \n',eq_nbr,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(problemcol(ii)))
else
eq_nbr=problemrow(ii)-aux_eq_nbr;
orig_var_index=M_.aux_vars(1,problemcol(ii)-M_.orig_endo_nbr).orig_index;
fprintf('Derivative of Equation %d with respect to Variable %s (initial value of %s: %g) \n',eq_nbr,deblank(M_.endo_names(orig_var_index,:)),deblank(M_.endo_names(orig_var_index,:)),x(problemcol(ii)))
end
end
end
end
@ -134,5 +134,5 @@ elseif strcmp(type,'static')
fprintf('%s and evaluating it at the steady state then results in a division by 0.\n',caller_string)
fprintf('%s If you are using model-local variables (# operator), check their values as well.\n',caller_string)
else
error('Unknown Type')
error('Unknown Type')
end

6
matlab/distributions/beta_specification.m
View File

@ -2,13 +2,13 @@ function [a, b] = beta_specification(mu, sigma2, lb, ub, name) % --*-- Unitary
% Returns the hyperparameters of the beta distribution given the expectation and variance.
%
% INPUTS
% INPUTS
% - mu [double] Expectation of the Gamma random variable.
% - sigma2 [double] Variance of the Gamma random variable.
% - lb [double] Lower bound of the domain (default is zero).
% - ub [double] Upper bound of the domain (default is one).
%
% OUTPUTS
% OUTPUTS
% - a [double] First hyperparameter of the Beta density.
% - b [double] Second hyperparameter of the Beta density.
@ -63,7 +63,7 @@ if sigma2>(1-mu)*mu
error('Beta prior%s. Given the declared prior expectation, prior lower and upper bounds, the prior std. has to be smaller than %f.',name2,sqrt((1-mu)*mu))
end
a = (1-mu)*mu*mu/sigma2-mu;
a = (1-mu)*mu*mu/sigma2-mu;
b = a*(1/mu-1);
%@test:1

16
matlab/distributions/compute_prior_mode.m
View File

@ -1,8 +1,8 @@
function m = compute_prior_mode(hyperparameters,shape) % --*-- Unitary tests --*--
% This function computes the mode of the prior distribution given the (two, three or four) hyperparameters
% of the prior distribution.
%
% INPUTS
%
% INPUTS
% hyperparameters [double] 1*n vector of hyper parameters.
% shape [integer] scalar specifying the prior shape:
% shape=1 => Beta distribution,
@ -12,14 +12,14 @@ function m = compute_prior_mode(hyperparameters,shape) % --*-- Unitary tests --*
% shape=5 => Uniform distribution,
% shape=6 => Inverse Gamma (type 2) distribution,
% shape=8 => Weibull distribution.
%
% OUTPUTS
%
% OUTPUTS
% m [double] scalar or 2*1 vector, the prior mode.
%
% REMARKS
% [1] The size of the vector of hyperparameters is 3 when the Gamma or Inverse Gamma is shifted and 4 when
% the support of the Beta distribution is not [0,1].
% [2] The hyperparameters of the uniform distribution are the lower and upper bounds.
% REMARKS
% [1] The size of the vector of hyperparameters is 3 when the Gamma or Inverse Gamma is shifted and 4 when
% the support of the Beta distribution is not [0,1].
% [2] The hyperparameters of the uniform distribution are the lower and upper bounds.
% [3] The uniform distribution has an infinity of modes. In this case the function returns the prior mean.
% [4] For the beta distribution we can have 1, 2 or an infinity of modes.

4
matlab/distributions/gamma_specification.m
View File

@ -2,13 +2,13 @@ function [a, b] = gamma_specification(mu, sigma2, lb, name) % --*-- Unitary te
% Returns the hyperparameters of the gamma distribution given the expectation and variance.
%
% INPUTS
% INPUTS
% - mu [double] Expectation of the Gamma random variable.
% - sigma2 [double] Variance of the Gamma random variable.
% - lb [double] Lower bound of the domain (default is zero).
% - name [string] Name of the parameter (or random variable).
%
% OUTPUTS
% OUTPUTS
% - a [double] First hyperparameter of the Gamma density (shape).
% - b [double] Second hyperparameter of the Gamma density (scale).

6
matlab/distributions/inverse_gamma_specification.m
View File

@ -2,7 +2,7 @@ function [s,nu] = inverse_gamma_specification(mu, sigma2, lb, type, use_fzero_fl
% Computes the inverse Gamma hyperparameters from the prior mean and standard deviation.
%
% INPUTS
% INPUTS
% - mu [double] scalar, prior mean.
% - sigma2 [double] positive scalar, prior variance.
% - type [integer] scalar equal to 1 or 2, type of the inverse gamma distribution
@ -10,11 +10,11 @@ function [s,nu] = inverse_gamma_specification(mu, sigma2, lb, type, use_fzero_fl
% dynare's implementation of the secant method otherwise.
% - name [string] name of the parameter or random variable.
%
% OUTPUS
% OUTPUS
% - s [double] scalar, first hyperparameter.
% - nu [double] scalar, second hyperparameter.
%
% REMARK
% REMARK
% The call to the matlab's implementation of the secant method is here for testing purpose and should not be used. This routine fails
% more often in finding an interval for nu containing a signe change because it expands the interval on both sides and eventually
% violates the condition nu>2.

28
matlab/distributions/lpdfgweibull.m
View File

@ -1,14 +1,14 @@
function [ldens,Dldens,D2ldens] = lpdfgweibull(x,a,b,c) % --*-- Unitary tests --*--
% Evaluates the logged Weibull PDF at x.
% Evaluates the logged Weibull PDF at x.
%
% INPUTS
% INPUTS
% - x [double] m*n matrix of points where the (logged) density will be evaluated,
% - a [double] m*n matrix of First Weibull distribution parameters (shape parameter, k),
% - b [double] m*n matrix of Second Weibull distribution parameters (scale parameter, λ),
% - c [double] m*n matrix of Third Weibull distribution parameters (location parameter, default is 0).
%
% OUTPUTS
% OUTPUTS
% - ldens [double] m*n matrix of logged (generalized) Weibull densities.
% - Dldens [double] m*n matrix (first order derivatives w.r.t. x)
% - D2ldens [double] m*n matrix (second order derivatives w.r.t. x)
@ -54,11 +54,11 @@ end
if ~isscalar(x)
if isscalar(a)
a = repmat(a, size(x));
end
end
if isscalar(b)
b = repmat(b, size(x));
end
if isscalar(c)
if isscalar(c)
c = repmat(c, size(x));
end
end
@ -77,7 +77,7 @@ if isempty(idx), return, end
jdx = find( abs(a-1)<1e-12 & x>=c & (x-c)<1e-12) ;
ldens(jdx) = 1.0;
if ~isempty(idx)
x0 = x(idx)-c(idx);
x1 = x0./b(idx);
@ -136,7 +136,7 @@ end
%$ catch
%$ t(1) = false;
%$ end
%$
%$
%$ if t(1)
%$ t(2) = isinf(a);
%$ end
@ -179,7 +179,7 @@ end
%$ scale = 1;
%$ shape = 2;
%$ mode = scale*((shape-1)/shape)^(1/shape);
%$
%$
%$ try
%$ [a, b, c] = lpdfgweibull(mode, shape, scale);
%$ p = rand(1000,1)*4;
@ -203,7 +203,7 @@ end
%$ scale = 1;
%$ shape = 2;
%$ density = @(x) exp(lpdfgweibull(x,shape,scale));
%$
%$
%$ try
%$ if isoctave
%$ s = quadv(density, .0000000001, 100000,1e-10);
@ -226,7 +226,7 @@ end
%$ scale = 1;
%$ shape = 1;
%$ density = @(x) exp(lpdfgweibull(x,shape,scale));
%$
%$
%$ try
%$ if isoctave
%$ s = quadv(density, .0000000001, 100000,1e-10);
@ -249,7 +249,7 @@ end
%$ scale = 1;
%$ shape = .5;
%$ density = @(x) exp(lpdfgweibull(x,shape,scale));
%$
%$
%$ try
%$ if isoctave
%$ s = quadv(density, .0000000001, 100000,1e-10)
@ -276,7 +276,7 @@ end
%$ scale = 1;
%$ shape = 2;
%$ xdens = @(x) x.*exp(lpdfgweibull(x,shape,scale));
%$
%$
%$ try
%$ if isoctave
%$ s = quadv(xdens, .0000000001, 20,1e-10)
@ -299,7 +299,7 @@ end
%$ scale = 1;
%$ shape = 1;
%$ xdens = @(x) x.*exp(lpdfgweibull(x,shape,scale));
%$
%$
%$ try
%$ if isoctave
%$ s = quadv(xdens, .0000000001, 100000,1e-10)
@ -322,7 +322,7 @@ end
%$ scale = 1;
%$ shape = .5;
%$ xdens = @(x) x.*exp(lpdfgweibull(x,shape,scale));
%$
%$
%$ try
%$ if isoctave
%$ s = quadv(xdens, .0000000001, 100000,1e-10)

16
matlab/distributions/mode_and_variance_to_mean.m
View File

@ -1,21 +1,21 @@
function [mu, parameters] = mode_and_variance_to_mean(m,s2,distribution,lower_bound,upper_bound)
% This function computes the mean of a distribution given the mode and variance of this distribution.
%
% INPUTS
% INPUTS
% m [double] scalar, mode of the distribution.
% s2 [double] scalar, variance of the distribution.
% distribution [integer] scalar for the distribution shape
% distribution [integer] scalar for the distribution shape
% 1 gamma
% 2 inv-gamma-2
% 3 inv-gamma-1
% 4 beta
% 4 beta
% lower_bound [double] scalar, lower bound of the random variable support (optional).
% upper_bound [double] scalar, upper bound of the random variable support (optional).
%
% OUTPUT
%
% OUTPUT
% mu [double] scalar, mean of the distribution.
% parameters [double] 2*1 vector, parameters of the distribution.
%
%
% Copyright (C) 2009 Dynare Team
%
@ -34,7 +34,7 @@ function [mu, parameters] = mode_and_variance_to_mean(m,s2,distribution,lower_bo
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
% Check input aruments.
% Check input aruments.
if ~(nargin==3 || nargin==5 || nargin==4 )
error('mode_and_variance_to mean:: 3 or 5 input arguments are needed!')
end
@ -80,7 +80,7 @@ if (distribution==1)% Gamma distribution
end
if (m-lower_bound)<1e-12
error('The gamma distribution should be specified with the mean and variance.')
end
end
m = m - lower_bound ;
beta = -.5*m*(1-sqrt(1+4*s2/(m*m))) ;
alpha = (m+beta)/beta ;

16
matlab/distributions/multivariate_normal_pdf.m
View File

@ -1,18 +1,18 @@
function density = multivariate_normal_pdf(X,Mean,Sigma_upper_chol,n)
% Evaluates the density of a multivariate gaussian, with expectation Mean
% and variance Sigma_upper_chol'*Sigma_upper_chol, at X.
%
%
% INPUTS
%
% X [double] 1*n vector
% INPUTS
%
% X [double] 1*n vector
% Mean [double] 1*n vector, expectation of the multivariate random variable.
% Sigma_upper_chol [double] n*n matrix, upper triangular Cholesky decomposition of Sigma (the covariance matrix).
% n [integer] dimension.
%
% OUTPUTS
% density [double] density
%
%
% OUTPUTS
% density [double] density
%
% SPECIAL REQUIREMENTS
% Copyright (C) 2003-2017 Dynare Team
@ -31,6 +31,6 @@ function density = multivariate_normal_pdf(X,Mean,Sigma_upper_chol,n)
%
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
density = (2*pi)^(-.5*n) * ...
density = (2*pi)^(-.5*n) * ...
prod(diag(Sigma_upper_chol))^(-1) * ...
exp(-.5*(X-Mean)*(Sigma_upper_chol\(transpose(Sigma_upper_chol)\transpose(X-Mean))));

12
matlab/distributions/multivariate_student_pdf.m
View File

@ -2,16 +2,16 @@ function density = multivariate_student_pdf(X,Mean,Sigma_upper_chol,df)
% Evaluates the density of a multivariate student, with expectation Mean,
% variance Sigma_upper_chol'*Sigma_upper_chol and degrees of freedom df, at X.
%
% INPUTS
% INPUTS
%
% X [double] 1*n vector
% X [double] 1*n vector
% Mean [double] 1*n vector, expectation of the multivariate random variable.
% Sigma_upper_chol [double] n*n matrix, upper triangular Cholesky decomposition of Sigma (the "covariance matrix").
% df [integer] degrees of freedom.
%
% OUTPUTS
% density [double] density.
%
%
% OUTPUTS
% density [double] density.
%
% SPECIAL REQUIREMENTS
% Copyright (C) 2003-2017 Dynare Team

8
matlab/distributions/rand_inverse_wishart.m
View File

@ -17,12 +17,12 @@ function G = rand_inverse_wishart(m, v, H_inv_upper_chol)
% OUTPUTS:
% G: G ~ IW(m, v, H) where H = inv(H_inv_upper_chol'*H_inv_upper_chol)
% or, equivalently, using the correspondence between Wishart and
% inverse-Wishart: inv(G) ~ W(m, v, S) where
% inverse-Wishart: inv(G) ~ W(m, v, S) where
% S = H_inv_upper_chol'*H_inv_upper_chol = inv(H)
%
%
% SPECIAL REQUIREMENT
% none
%
%
% Copyright (C) 2003-2009 Dynare Team
%
@ -41,7 +41,7 @@ function G = rand_inverse_wishart(m, v, H_inv_upper_chol)
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
X = randn(v, m) * H_inv_upper_chol;
X = randn(v, m) * H_inv_upper_chol;
% At this point, X'*X is Wishart distributed

Some files were not shown because too many files have changed in this diff Show More