It's hard to tell. The estimate of sigma_u is usually obtained by
subtracting within sum of squares from the overall sum of squares, and
it is not guaranteed that the result is non-negative. Wooldridge
discusses this stuff extensively in his "Cross Section and Panel Data"
book.
Conceptually, this may also mean that the estimate of variance is on
the boundary of the admissible space [0,+infty), and testing on the
boundary is a whole area. There's a little bit about it in -help
j_chibar- that refers to an STB publication. The derivatives at this
max do not have to be zeroes, which must be the reason Wald test is
not applicable.
Now, why does this arise? The first answer is that this may be due to
zero (or rather small) variance of the individual effects. If you
generate white noise and fit a panel model to it, in half of the cases
the estimate of sigma will be positive, and half of the times it will
be restricted to zero.
The second answer is that there might be some specification problems
with your model and/or your data. It is usually impossible to tell
what exactly is wrong... except that something IS wrong.
Stas
On Fri, 24 Sep 2004 18:09:50 -0400, Jeannette Wicks-Lim
<[email protected]> wrote:
> Hi,
>
> I am trying to run the Hausman test to compare the estimates from XTREG, FE
> to XTREG, RE. But, when I use the command "xthaus" I get the following
> message:
>
> . xthaus
> Estimate of sigma_u = 0, random-effects estimator has degenerated to pooled
> OLS and the Wald test from xthausman may not be appropriate. See [R]
> hausman
> for a more general implementation of the Hausman test.
>
> Can someone explain to me why might my sigma_u = 0? Does it indicate that
> there is something wrong with the way I've specified my model or that
> there's something wrong with the data I'm using (not enough variation?)?
>
> Thanks so much,
> Jeannette
>
> *
> * 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/
>
--
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/
