Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: fixed effect and random effect models


From   "Gary Tian" <gtian@uow.edu.au>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: fixed effect and random effect models
Date   Mon, 22 Oct 2007 18:36:11 +1000

Dear Stata users,

 

This question might have already been asked and answered here. I have a
problem in choosing between pooled regression, fixed and random. The problem
might be too simple to ask but your answer would be highly appreciated.

I have unbalanced firm data: 50 firms, some firms have long data than
others. I have run a simple Lintener model with change in dividend as
function of earnings and dividend lag. 

(1) ASSUME THE x'S were independent of the omitted fixed effects; the
Breusch-Pagan test (Greene pp. 298-299) would make sense. The result does
not reject, so prefer pooled regression.

(2) Then I also run Hausman test to see whether fixed effect model might be
a right one:

(2.1) compare fixed with random: indicating rejecting the consistent under
null, so prefer the fixed.

(2.2) also compare fixed with pooled regression: also reject the pooled
regression.

However, the fixed effect model itself has insignificant F test that all
u_i=0, P-value is bigger than even 10%. 
 

Regards,

 

Gary

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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