Page 1 of 1
Heteroscedasticity in VAR
Posted: Fri Oct 30, 2009 2:56 pm
by weber
Dear All:
To my knowledge, Eviews runs VAR (Vector Autoregressive Model) by an OLS approach. I would like to know if Eviews can do a VAR that deals with heteroscedasticity (for example, uses WLS) or if Eviews can do a VAR based on GMM?
If yes, could you please let me know how to do them? Many thanks
Best,
Weber
Re: Heteroscedasticity in VAR
Posted: Fri Oct 30, 2009 5:02 pm
by EViews Glenn
You could create a system from the VAR specification and run the estimation using one of the methods supported there...
Re: Heteroscedasticity in VAR
Posted: Fri Oct 30, 2009 8:32 pm
by weber
Thank you very much for your quick response.
However, I think I probably have not put the question very clearly. Here is what I want: I need to estimate the VAR itself by WLS or GMM, and generate the following impulse response functions.
The “system objects” (from VAR) you have proposed look like just to extract the equation formats from VAR and not to help estimate the VAR itself. I might be wrong.
Please help, thanks a lot.
Weber
Re: Heteroscedasticity in VAR
Posted: Mon Nov 02, 2009 9:25 am
by EViews Glenn
You can use the system object that has the VAR specification to estimate the parameters of the specification using a variety of methods (including various methods which admit heteroskedasticity). Click on the estimate button.
There are not, however, built-in impulse response tools in the system so you'll have to roll your own for that part.
Re: Heteroscedasticity in VAR
Posted: Thu Nov 12, 2009 7:47 am
by donihue
There are not, however, built-in impulse response tools in the system so you'll have to roll your own for that part.
Would you have any examples of an "own-rolled" code for the IRFs for a VAR estimated via the System command? They would be a very big help!
Regards
Donihue