EViews.com

Hello again, Esther.

Sorry to be such a pest, but I am having some trouble with the models generated by your excellent BVAR add-in.

As you will see from the attached screen-shot, in some cases (this is one) the generated model does not at all match the estimated model. Perhaps this is because I have such long names?

Also, I added a variable in the "exogenous" box, but it did not appear in the estimated model; perhaps I have done something wrong?

Regards
Donihue

Could you provide the workfile too?

Gareth,

As the workfile for the example I sent is very complicated, I have attached a much simpler one, which illustrates the same problem.

In the attached file, "var01" is the straight VAR, including an exogenous variable; "var02" is the same VAR without the exogenous variable.

Next, "bvar" is the flat-flat BVAR obtained from the Add-in. It was specified to include an exogenous variable, so as to match "var01", but in fact does not; its coefficients instead match those of "var02".

Finally, "ff1" is the model generated by the Add-in. As you can see, despite the short names, the text representation does not at all match the true VAR - it appears that the model picks up the right coefficients, but applies them to the wrong variables.

I imagine that this is very easy to fix, but I don't have enough programming skills to do so.

Thanks for your help.

Regards
Donihue

The "Model Option" is updated.

Please note that the current version of BVAR allows for exogenous variables only for the Sims-Zha model. Therefore, exogenous variables cannot be included in the diffuse prior model.
However, as you know, the estimates of the diffuse prior model are identical to those in the (classical) VAR model which allows for exogenous variables.

Thanks very much, Esther, for your quick response.

I downloaded just now the BVAR add-in and re-installed it, then re-ran my models. Unfortunately, I still find the same problem with the model coefficients - see attached workfile.

Could it be that the version of the Add-in on your web-site is not the one you revised?

Regards
Donihue

The one that was uploaded on Friday addressed the exogenous variables in flat-flat priors only. We'll have a fix for the model part in a few days.

Many thanks!

Regards
Donihue

Just put a new one up.

Thanks very much, Gareth. I have tested it and it seems to work correctly.

Regards
Donihue

BTW, is there any technical reason why it is not possible to add exogenous variables when using the flat-flat prior?

Nothing more technical than it is complicated to fit into the Add-in, and, given that it offers nothing that the built in VAR routines do, not worth it.

Yes, of course. I understand fully.

Regards
Donihue