Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st:problem of using Hausman test :fixed effect vs. random effect


From   Mandy fu <mandy.fu1@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st:problem of using Hausman test :fixed effect vs. random effect
Date   Fri, 24 Jul 2009 19:57:21 -0400

Hi all,

I was wondering if any one could give me some advice of choosing panel
data estimation methods, such as fixed effect(FE) and random
effect(RE). Sorry that my question is more related to basic
statistical knowledge instead of Stata. I thought perhaps Statistics
Experts here could give me some helpful comments.

The typical way in textbook of testing between FE and RE is Hausman
test. But it is said that Hausman test almost always rejects RE. My
own experience also proves this:without one exception,  all the panel
data sets I have worked on indicate that Hausman test shows FE is
appropriate compared to RE.

This should not be a problem when FE and RE main results are
close,because I can say, "I checked both FE and RE, although Hausman
test shows FE is more appropriate. The results are similar using RE
and FE."

But , my question is, what if main results from FE and RE are not
close? Is there any better criteria  than Hausman test to choose a
method between FE and RE?

Thanks a lot!

Mandy
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index