Bayesian VAR

For questions about EViews Add-ins available from the EViews Add-ins webpage. Note each add-in available on our webpage will have its own individual thread.

Moderators: EViews Gareth, EViews Moderator, EViews Esther

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: Bayesian VAR

Postby EViews Esther » Thu Jan 20, 2011 9:07 am

Yes. It should say "relative cross-variable weight".

Thank you for your attention.

Sanke Charmer
Posts: 3
Joined: Sun Nov 07, 2010 2:10 pm

Re: Bayesian VAR

Postby Sanke Charmer » Thu Jan 20, 2011 9:29 am

Thanks to your for your time, Esther.

I'm having another problems with the last version of the BVAR add-inn. I'm not sure what am I doing wrong:

1) Is there a limit in the list of endogenous variables on the "Bayesian VAR" dialog? Because I can't put more than 50 characters there.

2) The error message "_YLAG01 is not defined" appears often. In fact I've unable to estimate my BVAR with the Koko's priors - I've tried the default values of the priors, my own priors, and also diffent forecasting options.

3) The "model" option is great, but sometimes when the model is too long appears a "Syntax error" in the model object. It seems that the code doesnt fill the complete BVAR equation in the model source edit command line.

Thanks again, the BVAR is a great improvement in Eviews, as, to my knowledge, is the first step of Eviews in the promissory field of bayesian statistics.

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: Bayesian VAR

Postby EViews Esther » Thu Jan 20, 2011 5:57 pm

Sanke Charmer wrote:1) Is there a limit in the list of endogenous variables on the "Bayesian VAR" dialog? Because I can't put more than 50 characters there.

There is no maximum limit on the number of endogenous variables you can use. I make "the list of endogenous variables" box a bit bigger for your convenience. However, if you have many endogenous variables, the command line could be useful.

Sanke Charmer wrote:2) The error message "_YLAG01 is not defined" appears often. In fact I've unable to estimate my BVAR with the Koko's priors - I've tried the default values of the priors, my own priors, and also diffent forecasting options.

I fixed the problem.

donihue
Posts: 135
Joined: Wed Oct 07, 2009 8:51 am

Re: Bayesian VAR

Postby donihue » Fri Jan 21, 2011 4:46 am

Thank you for fixing the IRF and other problems, Esther. It is a great pleasure to work with someone as responsive as you.
Now, if only you could get that pesky "Schorfheide - del Negro DSGE prior" option built into your code, it would be perfect!

Regards
Donihue

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13305
Joined: Tue Sep 16, 2008 5:38 pm

Re: Bayesian VAR

Postby EViews Gareth » Mon Aug 08, 2011 7:51 am

Which is why we added it :D
Follow us on Twitter @IHSEViews

angelsky1207
Posts: 7
Joined: Sat Jul 16, 2011 8:26 am

Re: Bayesian VAR

Postby angelsky1207 » Thu Aug 11, 2011 1:35 pm

Hi i am a new learner of Bayesian VARs. The add-in on Eviews seems very useful. Is there any manual for this so that i can know better about the prior distributions? For the koko-minnesota prior, what is the overall tightness? And when i choose direct 1-step forecast, the coefficients are different from the coefficients under iternated forecast, is that right? sorry for the silly questions, but i am really not clear...

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 13305
Joined: Tue Sep 16, 2008 5:38 pm

Re: Bayesian VAR

Postby EViews Gareth » Thu Aug 11, 2011 1:52 pm

As with all Add-ins, the BVAR add-in comes with documentation. You can get to the documentation by clicking on the Docs button from the Manage Add-ins dialog.
Follow us on Twitter @IHSEViews

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: Bayesian VAR

Postby EViews Esther » Thu Aug 11, 2011 2:00 pm

angelsky1207 wrote: For the koko-minnesota prior, what is the overall tightness? And when i choose direct 1-step forecast, the coefficients are different from the coefficients under iternated forecast, is that right?

Regarding upon the Ko-Ko Minnesota prior, the first parameter "overall tightness" implies the scale (value) of the error covariance.
As mentioned in the documentation (p.9), Ko-Ko models will let you specify the model in different ways.

angelsky1207
Posts: 7
Joined: Sat Jul 16, 2011 8:26 am

Re: Bayesian VAR

Postby angelsky1207 » Tue Aug 30, 2011 6:57 am

EViews Esther wrote:
angelsky1207 wrote: For the koko-minnesota prior, what is the overall tightness? And when i choose direct 1-step forecast, the coefficients are different from the coefficients under iternated forecast, is that right?

Regarding upon the Ko-Ko Minnesota prior, the first parameter "overall tightness" implies the scale (value) of the error covariance.
As mentioned in the documentation (p.9), Ko-Ko models will let you specify the model in different ways.

Hi,
Thank you for your quick reply. I have read the documentation. But i am still confused that when i choose the forecast period =1, the interated forecast is not equal to the direct 1-step forecast, which in my opinion, these two forecasts should be the same for one-period forecast. And i also run BVAR without selection of forecast, the coefficients shown are different . Can you tell me why?

thank you again for your help.

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: Bayesian VAR

Postby EViews Esther » Tue Aug 30, 2011 9:34 am

There is an excellent paper to help your understanding: M. Marcellino, J. H. Stock, M. W. Watson (2005) "A comparison of direct and iterated multistep AR methods for forecasting macroeconomic time series", J. of Econometrics (http://www.economics.harvard.edu/facult ... _hstep.pdf).

Please let me know whether this paper would be helpful.

nlarse2
Posts: 1
Joined: Wed Jun 01, 2011 9:27 am

Re: Bayesian VAR

Postby nlarse2 » Thu Sep 01, 2011 11:55 am

Is there any BVAR addin for Eviews 6?

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: Bayesian VAR

Postby startz » Thu Sep 01, 2011 12:20 pm

All add-ins require version 7 or higher.

angelsky1207
Posts: 7
Joined: Sat Jul 16, 2011 8:26 am

Re: Bayesian VAR

Postby angelsky1207 » Mon Sep 05, 2011 4:40 am

EViews Esther wrote:There is an excellent paper to help your understanding: M. Marcellino, J. H. Stock, M. W. Watson (2005) "A comparison of direct and iterated multistep AR methods for forecasting macroeconomic time series", J. of Econometrics (http://www.economics.harvard.edu/facult ... _hstep.pdf).

Please let me know whether this paper would be helpful.

Hi,
i have read that paper. From that paper, we can tell that if the forecast horizon is 1, the interated equation and the direct forecast equation are the same except the error term. But when i try on eviews, the coefficients are different.

EViews Esther
EViews Developer
Posts: 149
Joined: Fri Sep 03, 2010 7:57 am

Re: Bayesian VAR

Postby EViews Esther » Fri Sep 09, 2011 4:24 pm

The aforementioned paper will give you the same model specification when h=1.

However, following Professor Gary Koop's idea (http://personal.strath.ac.uk/gary.koop/ ... bilis.html), the model specification in this addin would be different when implementing direct forecasts, compared to the specification when computing iterated forecasts. In particular, Koop writes the models as follows:

Iterated: Y(t) = A0 + Y(t-1)*A1 + .... + Y(t-p)*Ap + e(t)
Direct: Y(t+h) = A0 + Y(t)*A1 + .... + Y(t-(p-1))*Ap + e(t+h)

mjfl
Posts: 49
Joined: Mon Jul 18, 2011 8:18 am

Re: Bayesian VAR

Postby mjfl » Sat Oct 15, 2011 3:08 am

hi esther, can u let me know what i should change in the base code to allow a certain lag probably 4-7 instead of always starting from 1-4 in the var?
im using this

Code: Select all

'simulate with variable as shown
for !horizon= 95 to 122
smpl 1978q1 1978q1+!horizon
bvar(prior=3,m=a{!horizon},c) 4<---- can i put 4 7 instead of 1-4? z_piet8 i_disc

'forecast
smpl 1978q1+!horizon+1 1978q1+!horizon+1
a{!horizon}.solve(d=s)
series master_series = z_piet8_0
series ee= (z_piet8 - z_piet8_0)^2
next
smpl @all

'calculate
scalar r = sqr((1/28)*@sum(ee))
show r

hope u understand what im trying to do. have looked at the base codes but lots of syntax error when complied. pls help.


Return to “Add-in Support”

Who is online

Users browsing this forum: No registered users and 28 guests