Page 1 of 1

BVAR Eviews8 Forecast

Posted: Thu May 21, 2015 11:40 am
by forestG
I am using Eviews8 for estimation of BVAR and I am trying to use these for forecasts. Is Eviews implementing the direct or iterated forecast for this? Also, is the value output the forecast mean or median? I have looked in the user guide but cant find an answer. I am using the solve function for this. Maybe that is wrong.


Thank you in advance.

Re: BVAR Eviews8 Forecast

Posted: Thu May 21, 2015 12:05 pm
by EViews Gareth
The forecast is implemented in the classical sense - EViews just ignores the fact that the coefficients were estimated using Bayesian methods, and forecasts the same way as it would a classical model.

Re: BVAR Eviews8 Forecast

Posted: Thu May 21, 2015 2:40 pm
by forestG
Thank you. I was also wondering how the priors are implemented. I have for example
smpl @first @last-!n-!f+!i
var el.bvar(prior=lit,mu1=1,l1=1,l2=0.9,l3=20) 1 2 variable1variable2 variable3
el.makemodel(mod4)

this gives me very differnet forecast compared to a matlab toolbox. Is the first equation set as variable1 =1* variable1(-1) + 0*variable2(-1) + 0*variable3(-1) or is it set differently maybe such as: variable1 =1* variable1(-1) + 1*variable2(-1) + 1*variable3(-1)

Re: BVAR Eviews8 Forecast

Posted: Fri May 22, 2015 1:56 pm
by EViews Esther
The variance of the coefficients is assumed to have the form with the *three* hyperparameters (please see the attachment).

Can you provide me your Matlab script and the name of the toolbox(es) required for the script? I would like to run the script to replicate the result.