Dev

dev@dynare.org

3402 discussions

Re: [DynareDev] options_.diffuse_d
by ondra.kamenik＠gmail.com 14 Feb '10

14 Feb '10

I agree. Ondra K. ------Original Message------ From: Michel Juillard To: List for Dynare developers To: Ondra Kamenik To: Benes, Jaromir Sent: Feb 14, 2010 14:21 Subject: options_.diffuse_d When we implemented the diffuse filter/smoother, we used options_.diffuse_d to record the number of iterations spent in the diffuse part of the filter. 1) in the filter functions in kalman/likelihood, it is not used anymore except in univariate_diffuse_kalman_filter_corr.m. It is still present in the old (and not yet replaced) smoother functions. 2) in the multivariate case, diffuse_d can be interpreted as the number of periods for which we use the diffuse step. In the univariate case, diffuse_d is incremented inside the variable loop and can't be interpreted this way 3) when there is no missing observations, the number of period iterations spent in the diffuse step is equal to the highest order of integration of the variables of the model. The field name in oo_.smoother reflect this. But hen missing observations are present, its interpretation is much less clear. 4) In the past, we wanted to monitor diffuse_d, to make sure that it didn't change because of numerical issues. Now that we are using the Schur transformation of the transition equation, the numerical instability seems to have disappeared. All that to say that my inclination would be for dropping oo_.diffuse_d entirely. What do you think? Ondra, I know that this will create problems for GPM and its current infrastructure code. But I think that we should handle this problem in a different way, taking into account the number of periods with missing observations at the beginning of the sample and the order of intergation of the GPM model (== 1) that we know by construction. All the best, Michel Sent from my BlackBerry® wireless device

1 0

Re: [DynareDev] [Dynare Git] Dynare branch, master, updated. a83c0e68c70df4481ab269a1de493eabb804df71
by Michel Juillard 12 Feb '10

12 Feb '10

I wonder whether ierating the product of the transition matrix in forecast is not less accurate than taking the power of the matrix. But I agree that it must be faster for large matrices. Maybe we should decompose the matrix before the loop and only take the power of the eigenvalues. Best Michel Stephane Adjemian wrote: > This is an automated email from the git hooks/post-receive script. It was > generated because a ref change was pushed to the repository containing > the project "Dynare". > > The branch, master has been updated > via a83c0e68c70df4481ab269a1de493eabb804df71 (commit) > from a828c3d2c1edf1bc4465561a425a9591910c0dc3 (commit) > > Those revisions listed above that are new to this repository have > not appeared on any other notification email; so we list those > revisions in full, below. > > - Log ----------------------------------------------------------------- > commit a83c0e68c70df4481ab269a1de493eabb804df71 > Author: Stéphane Adjemian (Charybdis) <stephane.adjemian(a)ens.fr> > Date: Fri Feb 12 12:31:32 2010 +0100 > > Bug fix and efficiency changes in smoothing routines. > > diff --git a/matlab/DiffuseKalmanSmoother1.m b/matlab/DiffuseKalmanSmoother1.m > index 8e624ad..ef64185 100644 > --- a/matlab/DiffuseKalmanSmoother1.m > +++ b/matlab/DiffuseKalmanSmoother1.m > @@ -59,7 +59,7 @@ spstar = size(Pstar1); > v = zeros(pp,smpl); > a = zeros(mm,smpl+1); > atilde = zeros(mm,smpl); > -aK = zeros(nk,mm,smpl+1); > +aK = zeros(nk,mm,smpl+nk); > iF = zeros(pp,pp,smpl); > Fstar = zeros(pp,pp,smpl); > iFinf = zeros(pp,pp,smpl); > @@ -71,11 +71,11 @@ P = zeros(mm,mm,smpl+1); > Pstar = zeros(spstar(1),spstar(2),smpl+1); Pstar(:,:,1) = Pstar1; > Pinf = zeros(spinf(1),spinf(2),smpl+1); Pinf(:,:,1) = Pinf1; > crit = options_.kalman_tol; > -crit1 = 1.e-8; > +crit1 = 1.e-8; > steady = smpl; > rr = size(Q,1); > QQ = R*Q*transpose(R); > -QRt = Q*transpose(R); > +QRt = Q*transpose(R); > alphahat = zeros(mm,smpl); > etahat = zeros(rr,smpl); > r = zeros(mm,smpl+1); > @@ -99,7 +99,7 @@ while rank(Pinf(:,:,t+1),crit1) & t<smpl > a(:,t+1) = T*atilde(:,t); > aK(1,:,t+1) = a(:,t+1); > for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Linf(:,:,t) = T - Kinf(:,:,t)*Z; > Fstar(:,:,t) = Pstar(mf,mf,t); > @@ -129,9 +129,9 @@ while notsteady & t<smpl > K(:,:,t) = T*PZI; > L(:,:,t) = T-K(:,:,t)*Z; > a(:,t+1) = T*atilde(:,t); > - aK(1,:,t+1) = a(:,t+1); > + aK(1,:,t+1) = a(:,t+1); > for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T)-T*P(:,mf,t)*transpose(K(:,:,t)) + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -154,7 +154,7 @@ while t<smpl > a(:,t+1) = T*a(:,t) + K_s*v(:,t); > aK(1,:,t+1) = a(:,t+1); > for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > end > t = smpl+1; > @@ -174,4 +174,4 @@ if d > alphahat(:,t) = a(:,t) + Pstar(:,:,t)*r0(:,t) + Pinf(:,:,t)*r1(:,t); > etahat(:,t) = QRt*r0(:,t); > end > -end > +end > \ No newline at end of file > diff --git a/matlab/DiffuseKalmanSmoother1_Z.m b/matlab/DiffuseKalmanSmoother1_Z.m > index f5614af..debf47c 100644 > --- a/matlab/DiffuseKalmanSmoother1_Z.m > +++ b/matlab/DiffuseKalmanSmoother1_Z.m > @@ -105,8 +105,8 @@ while rank(Pinf(:,:,t+1),crit1) & t<smpl > a(:,t+1) = T*atilde(:,t); > aK(1,:,t+1) = a(:,t+1); > % isn't a meaningless as long as we are in the diffuse part? MJ > - for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Linf(:,:,t) = T - Kinf(:,:,t)*Z; > Fstar(:,:,t) = Z*Pstar(:,:,t)*Z'; > @@ -138,10 +138,13 @@ while notsteady & t<smpl > L(:,:,t) = T-K(:,:,t)*Z; > a(:,t+1) = T*atilde(:,t); > Pf = P(:,:,t); > + aK(1,:,t+1) = a(:,t+1); > for jnk=1:nk, > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*atilde(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > P(:,:,t+1) = T*P(:,:,t)*T'-T*P(:,:,t)*Z'*K(:,:,t)' + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -162,10 +165,13 @@ while t<smpl > atilde(:,t) = a(:,t) + PZI*v(:,t); > a(:,t+1) = T*atilde(:,t); > Pf = P(:,:,t); > - for jnk=1:nk, > + aK(1,:,t+1) = a(:,t+1); > + for jnk=1:nk > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*atilde(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > end > t = smpl+1; > diff --git a/matlab/DiffuseKalmanSmoother3.m b/matlab/DiffuseKalmanSmoother3.m > index 1127cf6..514890c 100644 > --- a/matlab/DiffuseKalmanSmoother3.m > +++ b/matlab/DiffuseKalmanSmoother3.m > @@ -165,13 +165,14 @@ while newRank & t < smpl > Pstar(:,:,t)=tril(Pstar(:,:,t))+transpose(tril(Pstar(:,:,t),-1)); > end > end > - a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + a1(:,t+1) = T*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*transpose(T)+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*transpose(T); > - P0=Pinf(:,:,t+1); > + P0 = Pinf(:,:,t+1); > if newRank, > %newRank = any(diag(P0(mf,mf))>crit); > newRank = rank(P0,crit1); > @@ -207,12 +208,15 @@ while notsteady & t<smpl > P(:,:,t)=tril(P(:,:,t))+transpose(tril(P(:,:,t),-1)); > end > end > - a1(:,t+1) = T*a(:,t); > + a1(:,t+1) = T*a(:,t); > Pf = P(:,:,t); > + aK(1,:,t+1) = a1(:,t+1); > for jnk=1:nk, > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T) + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -239,10 +243,13 @@ while t<smpl > end > a1(:,t+1) = T*a(:,t); > Pf = P(:,:,t); > - for jnk=1:nk, > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=1:nk > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > end > ri=zeros(mm,1); > diff --git a/matlab/DiffuseKalmanSmoother3_Z.m b/matlab/DiffuseKalmanSmoother3_Z.m > index b30e3b3..c6fa2a2 100644 > --- a/matlab/DiffuseKalmanSmoother3_Z.m > +++ b/matlab/DiffuseKalmanSmoother3_Z.m > @@ -91,7 +91,6 @@ Ki = zeros(mm,pp,smpl); > Kstar = zeros(mm,pp,smpl_diff); > P = zeros(mm,mm,smpl+1); > P1 = P; > -aK = zeros(nk,mm,smpl+nk); > PK = zeros(nk,mm,mm,smpl+nk); > Pstar = zeros(spstar(1),spstar(2),smpl_diff+1); Pstar(:,:,1) = Pstar1; > Pinf = zeros(spinf(1),spinf(2),smpl_diff+1); Pinf(:,:,1) = Pinf1; > @@ -162,9 +161,10 @@ while newRank & t < smpl > Pstar(:,:,t)=tril(Pstar(:,:,t))+tril(Pstar(:,:,t),-1)'; > end > end > - a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + a1(:,t+1) = T*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squueze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*T'+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*T'; > @@ -205,11 +205,14 @@ while notsteady & t<smpl > end > end > a1(:,t+1) = T*a(:,t); > - Pf = P(:,:,t); > + Pf = P(:,:,t); > + aK(1,:,t+1) = a1(:,t+1) > for jnk=1:nk, > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > P(:,:,t+1) = T*P(:,:,t)*T' + QQ; > % notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > diff --git a/matlab/DiffuseKalmanSmootherH1.m b/matlab/DiffuseKalmanSmootherH1.m > index 362a7a6..8e16023 100644 > --- a/matlab/DiffuseKalmanSmootherH1.m > +++ b/matlab/DiffuseKalmanSmootherH1.m > @@ -98,8 +98,9 @@ while rank(Pinf(:,:,t+1),crit1) & t<smpl > atilde(:,t) = a(:,t) + PZI*v(:,t); > Kinf(:,:,t) = T*PZI; > a(:,t+1) = T*atilde(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(1,:,t+1) = a(:,t+1); > + for jnk=1:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Linf(:,:,t) = T - Kinf(:,:,t)*Z; > Fstar(:,:,t) = Pstar(mf,mf,t) + H; > @@ -118,25 +119,26 @@ Pinf = Pinf(:,:,1:d); > notsteady = 1; > while notsteady & t<smpl > t = t+1; > - v(:,t) = Y(:,t) - a(mf,t) - trend(:,t); > + v(:,t) = Y(:,t) - a(mf,t) - trend(:,t); > P(:,:,t)=tril(P(:,:,t))+transpose(tril(P(:,:,t),-1)); > if rcond(P(mf,mf,t) + H) < crit > - return > - end > + return > + end > iF(:,:,t) = inv(P(mf,mf,t) + H); > PZI = P(:,mf,t)*iF(:,:,t); > atilde(:,t) = a(:,t) + PZI*v(:,t); > K(:,:,t) = T*PZI; > L(:,:,t) = T-K(:,:,t)*Z; > a(:,t+1) = T*atilde(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(1,:,t+1) = a(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T)-T*P(:,mf,t)*transpose(K(:,:,t)) + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > end > if t<smpl > - PZI_s = PZI; > + PZI_s = PZI; > K_s = K(:,:,t); > iF_s = iF(:,:,t); > P_s = P(:,:,t+1); > @@ -151,8 +153,9 @@ while t<smpl > v(:,t) = Y(:,t) - a(mf,t) - trend(:,t); > atilde(:,t) = a(:,t) + PZI_s*v(:,t); > a(:,t+1) = T*atilde(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(1,:,t+1) = a(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > end > t = smpl+1; > diff --git a/matlab/DiffuseKalmanSmootherH1_Z.m b/matlab/DiffuseKalmanSmootherH1_Z.m > index bbcdfd9..65fcd44 100644 > --- a/matlab/DiffuseKalmanSmootherH1_Z.m > +++ b/matlab/DiffuseKalmanSmootherH1_Z.m > @@ -108,8 +108,8 @@ while rank(Pinf(:,:,t+1),crit1) & t<smpl > a(:,t+1) = T*atilde(:,t); > aK(1,:,t+1) = a(:,t+1); > % isn't a meaningless as long as we are in the diffuse part? MJ > - for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Linf(:,:,t) = T - Kinf(:,:,t)*Z; > Fstar(:,:,t) = Z*Pstar(:,:,t)*Z' + H; > @@ -141,10 +141,13 @@ while notsteady & t<smpl > L(:,:,t) = T-K(:,:,t)*Z; > a(:,t+1) = T*atilde(:,t); > Pf = P(:,:,t); > - for jnk=1:nk, > + aK(1,:,t+1) = a(:,t+1); > + for jnk=1:nk > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*atilde(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > P(:,:,t+1) = T*P(:,:,t)*T'-T*P(:,:,t)*Z'*K(:,:,t)' + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -165,10 +168,13 @@ while t<smpl > atilde(:,t) = a(:,t) + PZI*v(:,t); > a(:,t+1) = T*atilde(:,t); > Pf = P(:,:,t); > - for jnk=1:nk, > + aK(1,:,t+1) = a(:,t+1); > + for jnk=1:nk > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*atilde(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > end > t = smpl+1; > diff --git a/matlab/DiffuseKalmanSmootherH3.m b/matlab/DiffuseKalmanSmootherH3.m > index c7dc47f..afa98b5 100644 > --- a/matlab/DiffuseKalmanSmootherH3.m > +++ b/matlab/DiffuseKalmanSmootherH3.m > @@ -163,9 +163,10 @@ while newRank & t < smpl > Pstar(:,:,t)=tril(Pstar(:,:,t))+transpose(tril(Pstar(:,:,t),-1)); > end > end > - a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + a1(:,t+1) = T*a(:,t); > + aK(1,:,t+1) = a1(:,t+1) > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*transpose(T)+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*transpose(T); > @@ -206,8 +207,9 @@ while notsteady & t<smpl > end > end > a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T) + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -233,8 +235,9 @@ while t<smpl > end > end > a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > end > ri=zeros(mm,1); > diff --git a/matlab/DiffuseKalmanSmootherH3_Z.m b/matlab/DiffuseKalmanSmootherH3_Z.m > index c01b833..5eaf948 100644 > --- a/matlab/DiffuseKalmanSmootherH3_Z.m > +++ b/matlab/DiffuseKalmanSmootherH3_Z.m > @@ -91,7 +91,6 @@ L0 = zeros(mm,mm,pp,smpl_diff); > Kstar = zeros(mm,pp,smpl_diff); > P = zeros(mm,mm,smpl+1); > P1 = P; > -aK = zeros(nk,mm,smpl+nk); > PK = zeros(nk,mm,mm,smpl+nk); > Pstar = zeros(spstar(1),spstar(2),smpl_diff+1); Pstar(:,:,1) = Pstar1; > Pinf = zeros(spinf(1),spinf(2),smpl_diff+1); Pinf(:,:,1) = Pinf1; > @@ -163,9 +162,10 @@ while newRank & t < smpl > Pstar(:,:,t)=tril(Pstar(:,:,t))+tril(Pstar(:,:,t),-1)'; > end > end > - a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + a1(:,t+1) = T*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*T'+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*T'; > @@ -207,10 +207,13 @@ while notsteady & t<smpl > end > a1(:,t+1) = T*a(:,t); > Pf = P(:,:,t); > + aK(1,:,t+1) = a1(:,t+1); > for jnk=1:nk, > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)) ; > + end > end > P(:,:,t+1) = T*P(:,:,t)*T' + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -238,11 +241,14 @@ while t<smpl > end > end > a1(:,t+1) = T*a(:,t); > - Pf = P(:,:,t); > + Pf = P(:,:,t); > + aK(1,:,t+1) = a1(:,t+1); > for jnk=1:nk, > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > end > ri=zeros(mm,1); > diff --git a/matlab/DiffuseKalmanSmootherH3corr.m b/matlab/DiffuseKalmanSmootherH3corr.m > index 800362c..cfbd826 100644 > --- a/matlab/DiffuseKalmanSmootherH3corr.m > +++ b/matlab/DiffuseKalmanSmootherH3corr.m > @@ -122,8 +122,9 @@ while newRank & t < smpl > end > end > a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + aK(1,:,t+1) = a1(:,t+1) > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*transpose(T)+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*transpose(T); > @@ -159,8 +160,9 @@ while notsteady & t<smpl > end > end > a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T) + QQ; > notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > @@ -186,8 +188,9 @@ while t<smpl > end > end > a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > end > > diff --git a/matlab/missing_DiffuseKalmanSmoother1.m b/matlab/missing_DiffuseKalmanSmoother1.m > index da7d840..5d76505 100644 > --- a/matlab/missing_DiffuseKalmanSmoother1.m > +++ b/matlab/missing_DiffuseKalmanSmoother1.m > @@ -60,7 +60,7 @@ spstar = size(Pstar1); > v = zeros(pp,smpl); > a = zeros(mm,smpl+1); > atilde = zeros(mm,smpl); > -aK = zeros(nk,mm,smpl+1); > +aK = zeros(nk,mm,smpl+nk); > iF = zeros(pp,pp,smpl); > Fstar = zeros(pp,pp,smpl); > iFinf = zeros(pp,pp,smpl); > @@ -113,9 +113,9 @@ while rank(Pinf(:,:,t+1),crit1) & t<smpl > Pinf(:,:,t+1) = T*Pinf(:,:,t)*transpose(T)-T*Pinf(:,mf1,t)* ... > transpose(Kinf(:,di,t)); > end > - aK(1,:,t+1) = a(:,t+1); > - for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(1,:,t+1) = a(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > end > d = t; > @@ -150,7 +150,7 @@ while notsteady & t<smpl > end > aK(1,:,t+1) = a(:,t+1); > for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T)-T*P(:,mf,t)*transpose(K(:,:,t)) + QQ; > % notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > diff --git a/matlab/missing_DiffuseKalmanSmoother1_Z.m b/matlab/missing_DiffuseKalmanSmoother1_Z.m > index 097f1de..e640fa5 100644 > --- a/matlab/missing_DiffuseKalmanSmoother1_Z.m > +++ b/matlab/missing_DiffuseKalmanSmoother1_Z.m > @@ -121,7 +121,7 @@ while rank(Pinf(:,:,t+1),crit1) & t<smpl > aK(1,:,t+1) = a(:,t+1); > % isn't a meaningless as long as we are in the diffuse part? MJ > for jnk=2:nk, > - aK(jnk,:,t+jnk) = T^(jnk-1)*a(:,t+1); > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > end > d = t; > @@ -157,10 +157,13 @@ while notsteady & t<smpl > end > a(:,t+1) = T*atilde(:,t); > Pf = P(:,:,t); > - for jnk=1:nk, > + aK(1,:,t+1) = a(:,t+1); > + for jnk=1:nk > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*atilde(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > % notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > end > diff --git a/matlab/missing_DiffuseKalmanSmoother3.m b/matlab/missing_DiffuseKalmanSmoother3.m > index 8f2a041..c4c6df3 100644 > --- a/matlab/missing_DiffuseKalmanSmoother3.m > +++ b/matlab/missing_DiffuseKalmanSmoother3.m > @@ -91,11 +91,11 @@ Pinf = zeros(spinf(1),spinf(2),smpl_diff+1); Pinf(:,:,1) = Pinf1; > Pstar1 = Pstar; > Pinf1 = Pinf; > crit = options_.kalman_tol; > -crit1 = 1.e-6; > +crit1 = 1.e-6; > steady = smpl; > rr = size(Q,1); > QQ = R*Q*transpose(R); > -QRt = Q*transpose(R); > +QRt = Q*transpose(R); > alphahat = zeros(mm,smpl); > etahat = zeros(rr,smpl); > r = zeros(mm,smpl+1); > @@ -167,16 +167,17 @@ while newRank & t < smpl > Pstar(:,:,t)=tril(Pstar(:,:,t))+transpose(tril(Pstar(:,:,t),-1)); > end > end > - a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + a1(:,t+1) = T*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*transpose(T)+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*transpose(T); > P0=Pinf(:,:,t+1); > if newRank, > %newRank = any(diag(P0(mf,mf))>crit); > - newRank = rank(P0,crit1); > + newRank = rank(P0,crit1); > end > end > > @@ -212,10 +213,13 @@ while notsteady & t<smpl > end > a1(:,t+1) = T*a(:,t); > Pf = P(:,:,t); > + aK(1,:,t+1) = a1(:,t+1); > for jnk=1:nk, > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a1(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > P(:,:,t+1) = T*P(:,:,t)*transpose(T) + QQ; > % notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > diff --git a/matlab/missing_DiffuseKalmanSmoother3_Z.m b/matlab/missing_DiffuseKalmanSmoother3_Z.m > index 88cde2f..8a7313f 100644 > --- a/matlab/missing_DiffuseKalmanSmoother3_Z.m > +++ b/matlab/missing_DiffuseKalmanSmoother3_Z.m > @@ -92,7 +92,6 @@ Ki = zeros(mm,pp,smpl); > Kstar = zeros(mm,pp,smpl_diff); > P = zeros(mm,mm,smpl+1); > P1 = P; > -aK = zeros(nk,mm,smpl+nk); > PK = zeros(nk,mm,mm,smpl+nk); > Pstar = zeros(spstar(1),spstar(2),smpl_diff+1); Pstar(:,:,1) = Pstar1; > Pinf = zeros(spinf(1),spinf(2),smpl_diff+1); Pinf(:,:,1) = Pinf1; > @@ -164,9 +163,10 @@ while newRank & t < smpl > Pstar(:,:,t)=tril(Pstar(:,:,t))+tril(Pstar(:,:,t),-1)'; > end > end > - a1(:,t+1) = T*a(:,t); > - for jnk=1:nk, > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > + a1(:,t+1) = T*a(:,t); > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=2:nk > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > end > Pstar(:,:,t+1) = T*Pstar(:,:,t)*T'+ QQ; > Pinf(:,:,t+1) = T*Pinf(:,:,t)*T'; > @@ -209,10 +209,13 @@ while notsteady & t<smpl > end > a1(:,t+1) = T*a(:,t); > Pf = P(:,:,t); > - for jnk=1:nk, > + aK(1,:,t+1) = a1(:,t+1); > + for jnk=1:nk > Pf = T*Pf*T' + QQ; > - aK(jnk,:,t+jnk) = T^jnk*a(:,t); > PK(jnk,:,:,t+jnk) = Pf; > + if jnk>1 > + aK(jnk,:,t+jnk) = T*squeeze(aK(jnk-1,:,t+jnk-1)); > + end > end > P(:,:,t+1) = T*P(:,:,t)*T' + QQ; > % notsteady = ~(max(max(abs(P(:,:,t+1)-P(:,:,t))))<crit); > > ----------------------------------------------------------------------- > > Summary of changes: > matlab/DiffuseKalmanSmoother1.m | 16 ++++++++-------- > matlab/DiffuseKalmanSmoother1_Z.m | 16 +++++++++++----- > matlab/DiffuseKalmanSmoother3.m | 23 +++++++++++++++-------- > matlab/DiffuseKalmanSmoother3_Z.m | 15 +++++++++------ > matlab/DiffuseKalmanSmootherH1.m | 23 +++++++++++++---------- > matlab/DiffuseKalmanSmootherH1_Z.m | 18 ++++++++++++------ > matlab/DiffuseKalmanSmootherH3.m | 17 ++++++++++------- > matlab/DiffuseKalmanSmootherH3_Z.m | 20 +++++++++++++------- > matlab/DiffuseKalmanSmootherH3corr.m | 15 +++++++++------ > matlab/missing_DiffuseKalmanSmoother1.m | 10 +++++----- > matlab/missing_DiffuseKalmanSmoother1_Z.m | 9 ++++++--- > matlab/missing_DiffuseKalmanSmoother3.m | 18 +++++++++++------- > matlab/missing_DiffuseKalmanSmoother3_Z.m | 15 +++++++++------ > 13 files changed, 131 insertions(+), 84 deletions(-) > > > hooks/post-receive >

1 0

[Trac] #54: Add hp_filter
by Trac Dynare 10 Feb '10

10 Feb '10

#54: Add hp_filter --------------------------+------------------------------------------------- Reporter: stepan | Owner: stepan Type: feature | Status: new Priority: major | Milestone: 4.1 Component: Core M-files | Version: Keywords: | --------------------------+------------------------------------------------- Only theoretical hp filterd moments are available in the current version. We need to add a general function for hp filtering time series and call this new function when displaying simulated moements (with option hp filter). -- Ticket URL: <https://www.dynare.org/trac/ticket/54> Dynare <http://www.dynare.org> The Dynare project

1 2

Re: [DynareDev] [Dynare Git] Dynare branch, 4.1, updated. 4.1.0-30-g7d958d5
by Stéphane Adjemian 08 Feb '10

08 Feb '10

Hi Marco, I understand now. It's true that we have nothing for this. I don't remember the reason why an error message is issued in this case, instead of a warning message. I fix it as soon as possible by adding a trap in the diffuse filter files. Best, Stéphane. > thank you for spotting this. I am guilty for introducing that trap: the > reason was if the model is non-stationary and for some parameter > values the filtver gives the error : "there is not sufficient info to > initialize the filter", this is not trapped in dsgelikelihood and > therefore we get an error break. > Obviously, one solution would be to trap the error in dsgelikelihood, > but I found easier to that there, because it was the only place where > sometimes I got that error. > I'mnot sure if that error is now trapped or we still get the error break. > > Just to let you know what was behind there. > > best wishes, have a nice week end > Marco > -- > Marco Ratto, > Euro-area Economy Modelling Centre > (http://eemc.jrc.ec.europa.eu), > Joint Research Centre > The European Commission, > TP 361, 21027 ISPRA(VA), ITALY > Tel: +39 0332 78 5639 > Fax: +39 0332 78 5733 > marco.ratto(a)jrc.ec.europa.eu

2 1

Migration to Git complete
by Sébastien Villemot 01 Feb '10

01 Feb '10

Hi, The migration to Git is complete, the wiki page has been updated with installation instructions: http://www.dynare.org/DynareWiki/GitMigration The procedure has been tested on both Linux and Windows and works smoothly (though it is possible that the documentation still needs some clarifications). You need to send me an SSH key (as described on the page) to get write access to your personal public repository. As you have seen, a new mailing list has been created for receiving information about changes merged in the central repository. Have a nice week-end, -- Sébastien Villemot

2 2

Reminder: Migration to Git tomorrow
by Sébastien Villemot 28 Jan '10

28 Jan '10

Hi, Just to remind you that we plan to migrate to Git tomorrow. As a consequence, if you have uncommitted changes in your working copy, I encourage you to commit them tonight (provided they don't break anything). However, if you don't commit them tonight, you will still be able to copy them to your future git working copy. Tomorrow, when the migration is done, I will send instructions on the list on how to setup this on your computers. In the meantime, you can refer to the wiki page I created (see my previous mail). Best -- Sébastien Villemot

1 0

Re: [DynareDev] Estimation DLL design
by Stéphane Adjemian 27 Jan '10

27 Jan '10

Hi all, For the declaration of calibrated parameters we may use the estimated_params block by specifying priors with zero variance... Or we may add dirac delta function as a new prior shape. For the choice of the prior density I agree with Michel. The user has to choose the prior variance knowing that the weight of the prior (relative to the likelihood) will depend on the sub-sample size. Obviously we could imagine a way to automatically reweight the priors, but I don't think it would be a good idea. Best, Stéphane. Le mardi 26 janvier 2010 à 13:39 +0100, Michel Juillard a écrit : > Hi George, > > > > There are however few outstanding issues to be resolved. > > > > 1.a) Is Detrend data called only once as it looks like to me now? > No. in EstimationModule, function data_filtering is called a first time > and then several time in Loop on periods 1 to NP > > > I.e. can trend depend on different draws of parmaters and different > > steady states? > yes > > > > 1.b) Also, Can trend and the constant be different for different periods? > yes > > > > 2) Can there be fixed (calibrated) parameters' sets for different > > regimes/periods? > > > Good point. This should be permitted, but we don't have provision for > it. Neither an interface for declararing changing calibrated parameters, > nor an internal representation in the preprocessor or in Matlab. > We need to think about it some more, but the resolution of this problem > will only affect the design of parameter updates > > > 3) Passing SteadyState: > > Do we need to return SS from estimation? > yes > > If so, which one since each estimation period may eventually have its > > own different SS? > > > all of them. SS is then a matrix with as many columns as there are > regimes in the model > > 4) Should we not weight prior density for the length of the period the > > parameter and its prior are valid for? > > > I don't think so. The prior density describes believes about the > possible values of a parameter irrespective of the number of periods for > which this parameter matters. If the subsample where a given paramter > matters is long, the prior will have less influence on estimation as in > a subsample that is short. But it is what we want. > > All the best, > > Michel > >

1 0

Stable 4.1 branch
by Sébastien Villemot 25 Jan '10

25 Jan '10

Hi everyone, Just to remind you that, if you fix important bugs in the trunk, you also need to fix them yourselves in the 4.1 branch (using the "svn merge" command, until we migrate to git). It's difficult for me to decide which fixes should be merged, and how the merge should be done in case of conflict, so I prefer to let you make the decision when relevant. Best, -- Sébastien Villemot

1 0

New documentation for MSBVAR interface
by Tao Zha 21 Jan '10

21 Jan '10

Dear Michel: I have revised the MSBVAR interface documentation substantially. I attach a version here. Please work on this version. Thanks, Tao ------------------

1 0

Re: [DynareDev] [Trac] #9: Add derivatives of external functions in the model (was: Add derivatives of unknown functions in the model)
by Trac Dynare 19 Jan '10

19 Jan '10

#9: Add derivatives of external functions in the model ---------------------------+------------------------------------------------ Reporter: sebastien | Owner: houtanb Type: feature | Status: assigned Priority: major | Milestone: 4.2 Component: Preprocessor | Version: Resolution: | Keywords: ---------------------------+------------------------------------------------ Changes (by sebastien): * owner: sebastien => houtanb * status: accepted => assigned Old description: New description: An analysis is avaiable at: http://www.dynare.org/DynareWiki/ExternalFunctions -- -- Ticket URL: <https://www.dynare.org/trac/ticket/9#comment:5> Dynare <http://www.dynare.org> The Dynare project

1 0

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

Dev