I am trying to estimate a DSGE-VAR as in Del Negro and Schorfheide (2008). I have a doubt about the procedure. First, I estimate the DSGE model. Second, I generate simulated (dummy) observations from it. My problem is on which data I have to use in estimating the DSGE-VAR model. I set lambda at 0.75, so given that I estimate the DSGE model with 110 observations, the dummy observations should be 70.
Now, in estimating the DSGE-VAR I do not know which of the following options I should follow:
- Use only the 70 dummy observations
- Use both real observations and dummy ones, namely using 180 as whole
- Use only dummy observations, but re-scaled in some way
DSGE-VAR models are implemented in Dynare (since a while). This is not documented in the manual because there is (still) no specific dynare commands for this model, but I wrote a page on the wiki.
thanks a lot for your reply. I was already going throgh the wiki page and the dynare example you posted there. Nevertheless, they do not help me to clarify my doubt about which observations to use in the estimation of the DSGE-VAR: etiher real observations+dummy observations or only dummy observations.
You just have to provide real observations. Actually, Dynare does not generate dummy observations but uses theoretical second order moments for evaluating the likelihood associated to these dummy variables, as described on the wiki page or in the paper by del Negro and Schorfheide, otherwise we would add some noise here and the results would depend on the realizations of the dummy observations.
I am running a DSGE-VAR model under dynare 4.1.1. The model runs just fine untill it reaches the point where it generates the bayesian posterior IRFs, and I get the following error:
??? Error using ==> mtimes
Inner matrix dimensions must agree.
Error in ==> PosteriorIRF at 229
SIGMAtrOMEGA = SIGMAu_chol*OMEGAstar’;
Error in ==> dynare_estimation_1 at 1074
Error in ==> dynare_estimation at 62
Error in ==> ls2003_bvar at 199
Error in ==> dynare at 132
How can I solve this problem?