Page 1 of 1

Posterior conditional variance decomposition

PostPosted: Tue Feb 01, 2011 6:13 am
by TimRobinson
I would like to obtain a conditional variance decomposition at several horizons from the posterior of a model I have estimated. I have three issues:
(1) If I put conditional_variance_decomposition=2 into estimation Dynare falls over.
(2) If I put it in a stoch_simul command directly after estimation M_.sigma_e_is_diagonal is not defined in M so Dynare falls over.
(3) If after estimation I put M_.sigma_e_is_diagonal =1 in the mod file and then do stoch_simul with conditional_variance_decomposition=2 as an option it runs, but for what parameter values is the variance decomposition being done for?

Thanks,
Tim

Re: Posterior conditional variance decomposition

PostPosted: Tue Feb 01, 2011 2:05 pm
by jpfeifer
(1) As the manual says, conditional_variance_decomposition is a command only allowed for stoch_simul, not for estimation
(2) Which Dynare version are you using? Have you tried the newest snapshot, where some bugfixes regarding the variance decomposition have been made? The error message might be some bug in the earlier version
(3) The manual specifies that after running estimation, the parameter values are set to the posterior mode/ML estimates. Hence, the commands of stoch_simul, including the variance decomposition are performed at these new values.

Re: Posterior conditional variance decomposition

PostPosted: Wed Feb 02, 2011 7:49 am
by StephaneAdjemian
Hi,

Regarding (3), stoch_simul will consider a model calibrated at the posterior mode/ML if you do not run a metrpolis. If you run a metropolis parameters will be calibrated at the posterior mean.

Best, Stéphane.

Re: Posterior conditional variance decomposition

PostPosted: Tue Mar 27, 2012 10:10 am
by reubenpjacob
hi, where exactly is the conditional forecast error variance decomposition stored?
i used the command
stoch_simul(noprint, irf=41, nograph,conditional_variance_decomposition=[1 5 41]);

and looked for the related output in oo_
but i find no matrix titled oo_.conditional_variance_decomposition as the manual says. I wonder where I am going wrong.


Thanks
Reuben