Problem in producing IRFs within a Bayesian VAR (BVAR) framework, with the Original Minnesota Prior

[Maurine Bassem]

Hi everyone!

I was having a problem while trying to produce Impulse Response Functions (IRFs) within a BVAR framework. I am using the Original Minnesota Prior, with the option "varcov" to specify that I want to take the VAR covariance matrix as the prior's covariance matrix.

I am using the following code:

bayes, rseed (123) saving(Minn, replace) minnfixedcovprior( selftight(1) crosstight(1) varcov)  : var var1 var2 var3 var4, lags(1/1)     //to estimate a Bayesian var model

bayesvarstable      //tests the stability condition 

bayesirf create birf1, replace step (20) set (birfex2)          // to compute and store the bayes IRF results

Although the first 2 codes run smoothly, the last one gives me the following error:

 bayesirf create birf1, replace step (20) set (birfex2) 
(file birfex2.irf created)
(file birfex2.irf now active)
irfname birf1 not found in birfex2.irf
_c_var_model::var_irf():  3301  subscript invalid
                 <istmt>:     -  function returned error
r(3301);

What is even more confusing is that when I remove the option "varcov", the code runs smoothly, and when I change the prior (i.e. when I use the minniwishprior command instead of minnfixedcovprior), while keeping the option "varcov" the code still runs smoothly too.

I would be grateful if anyone could help me with this issue.

Thanks for your time!