Information Criteria

Dear all,

I’m estimating with Bayesian methods and I would be interested in how to get information criteria such as AIC or BIC. Unfortunately, I only found this post (Generalized Impulse response functions in BVAR in Dynare) on this topic and it’s not clear to me how to calculate them myself, because I’m not aware of how to get the “correct” likelihood for these formulas (such as those found in Koop (2003) Chapter 10.3.5).

As always, I would be very very grateful for any help!

You mean for the BVAR in Dynare?

No, for a state space model with unobserved components. (Sorry for the misunderstanding due to the linked topic!)

The likelihood would be the one depicted in Dynare after maximizing the likelihood with the statement:

Final value of minus the log posterior (or likelihood):

Of course, you need to be careful with the minus sign and need to account for the number of estimated parameters when computing the information criterion.