## Pooled Mean Group & Mean Group Estimation

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

Shin
Posts: 4
Joined: Sun Dec 25, 2016 5:41 pm

### Pooled Mean Group & Mean Group Estimation

Dear fellow friends,

I am currently running a model with pooled mean group.
However I do not find Hausman Test for validation of Pooled Mean and Mean Group. May I know what is the command of this test?
I have tried "Eq01.ranhaus" which Eq01 is estimated equation, however, it prompts error message.

Thanks a lot for your help.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 11901
Joined: Tue Sep 16, 2008 5:38 pm

### Re: Pooled Mean Group & Mean Group Estimation

EViews does not have this test built in.
Follow us on Twitter @IHSEViews

Shin
Posts: 4
Joined: Sun Dec 25, 2016 5:41 pm

### Re: Pooled Mean Group & Mean Group Estimation

Dear Gareth,

May I further understand how to evaluate which model is better then for PMG and MG?
Can we just skip this Hausman test and assume PMG is best to use?

Thank

lgehrig
Posts: 16
Joined: Fri Mar 17, 2017 8:10 am

### Re: Pooled Mean Group & Mean Group Estimation

Hey Shin

I've just stumbled across the same problem.
http://www.stata-journal.com/sjpdf.html ... num=st0125
Here they argue that just assuming PMG is the best model, is a bad idea. If the slopes are in fact not homogenous, but you are forcing them to be in the long-run equation, then the PMG estimator is not efficient.

The application of the Hausman test here relies on the fact that the long-run parameter estimates can be derived from the average of the country regressions (MG). This is consistent even under heterogeneity. But if slopes are in fact homogenous, PMG is more efficient.

The Hausman test is really easy to implement though. As shown for example in Applied Econometrics (Asteriou, Hall, 2007) (or see Wikipedia) it is just:

H=q' [var(q)]^(-1) q where q is the difference in the estimates. It is chi-square distributed with k dfs, where k is the number of coefficients in q.

Edit: actually - forget the "really easy". Its quite tedious work. While you can get the variance-covariance matrix for the PMG estimation quite easily, the MG variance covariance matrix is complicated. Note that in the above Stata Journal article they give the advice to use the same variance estimate (sigma squared) for both covariance matrices (to avoid non-positive definite covariance matrices).

Return to “Estimation”

### Who is online

Users browsing this forum: No registered users and 15 guests