Large Bayesian VAR

[KOBE24]

Hi Tom,

working on Peersman 2004 OBES I was wondering if the Bayesian VAR described in RATS is suited for handling a Large Dataset in the spirit of Banbura-Giannone-Reichlin Journal of Applied Econometrics 2010.
I had a quick look at the procedure: would it be possible implementing IRF Vand FEVD for, say, 30 series?

As usually,

thanks in advance.

[TomDoan]

Yes. It could handle that. You can't apply the full system Gibbs sampling that's used in the Peersman example because of the sheer size of the inversion required, but you can with the restricted form of prior that allows single equation methods.

[KOBE24]

Thanks a lot!

So I basically have to use the canmodel program described in chapter 10 of user's guide and then modifying Gibbs sampling adding the standard monte carlo procedure for IRF?

[jonasdovern]

Hi,
did you manage to estimate something along the lines of BGR2010? Are you willing to share the code here? Best regards, Jonas