# Re: st: Re: testing for cross-sectional independence

 From Stas Kolenikov To statalist@hsphsun2.harvard.edu Subject Re: st: Re: testing for cross-sectional independence Date Sat, 9 Oct 2004 12:21:35 -0400

```> >    I have an unbalanced panel for several hundred firms. the maximium
> > length is 7 whereas the short is 1 for some of the firms. the mean length is
> > 5.4.
> >
> >    I want to fit a fixed effect regression while one of my supervisor
> > prefers poolling regression. I don't know what kind of test can i use
> > to choose models between them
> >
> > can u give me some suggestions?

. use http://www.stata-press.com/data/r8/nlswork.dta
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. xtreg ln_wage union coll , fe i(id)

Fixed-effects (within) regression               Number of obs      =     19238
Group variable (i): idcode                      Number of groups   =      4150

R-sq:  within  = 0.0146                         Obs per group: min =         1
between = 0.0627                                        avg =       4.6
overall = 0.0412                                        max =        12

F(1,15087)         =    223.79
corr(u_i, Xb)  = 0.1228                         Prob > F           =    0.0000

------------------------------------------------------------------------------
ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
union |   .1106144   .0073942    14.96   0.000     .0961208     .125108
_cons |   1.728786   .0026265   658.22   0.000     1.723638    1.733934
-------------+----------------------------------------------------------------
sigma_u |  .42531648
sigma_e |  .27368425
rho |  .70717789   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0:     F(4149, 15087) =     7.75         Prob > F = 0.0000

The F-test is what you are looking for: are all of the fixed effects
equal to zero? If yes, we can pool the data. Here, we obviously cannot
do that.

Note that -xtreg, fe- will either explicitly or implicitly drop panels
with one observation only. If that is a concern for the sample to
become unrepresentative of the population, then you are bound to use
some sort of random effects procedures or multilevel analysis that
economists generally don't seem to have a lot of respect for.

--
Stas Kolenikov
http://stas.kolenikov.name
*
*   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/
```