Page 1 of 1

Auto-series in BVAR modelling

Posted: Mon Aug 05, 2019 2:24 am
by Mattdered
Hello,

I notice that we cannot use auto-series (e.g. log(x/y) if x and y are series) in BVARs, as I get this error when trying to simulate the IRFs.

Is there something fundamental that Eviews cannot handle about auto-series in the estimation of BVARs? Since if I estimate the BVAR using the auto-series and then the same BVAR method to estimate one with the equivalent series without auto-series (e.g x_l = log(x)), the coefficient/std error estimates are different in the two models?

(This is true in both EV9.5 and EV11).

Thanks.

Re: Auto-series in BVAR modelling

Posted: Mon Aug 05, 2019 3:28 am
by Mattdered
Apologies, forget about this. Issue was the way I was setting the sample before generating the series so estimation ended up with less observations.