Testing overidentifying SVAR models

[bonilla]

Hello,

I used montesvar.rpf procedure to estimate an overidentified SVAR (nvar=7 and nfree=20). I was wondering if it is possible to obtain the likelihood ratio test for the overidentified restrictions from the procedure. In fact, when using DMATRIX=MARGINALIZED, as in montesvar.rpf, we do not obtain the log likelihood restricted and unrestricted, which are neccesary for the likelihood ratio test. I think that the aswer is NO but, could I use DMATRIX=CONCENTRATED in order to get what I need without jeopardize the bayesian estimation? should I test my overindentified restriction separately?

Thanks a lot for your advices!!

[TomDoan]

Yes. Use DMATRIX=CONCENTRATED if you want to do an LR test. DMATRIX=MARGINALIZED computes the log marginal density, not the log likelihood.

[bonilla]

Thanks Tom!