On 9/4/07, avwilson <[email protected]> wrote:
> I have 2 questions, plus 2 follow ups.
> 1. is gllamm the right tool to use? And if not, what should I do?
There aren't many estimation commands in Stata that cannot be cast in
terms of -gllamm- :)). Whether that's the best tool to use is then the
question; probably not in your case.
> 2. if yes, then should I worry about the condition number and variances? And
> what could I do to improve on them?
I would say that either of the results suggest absence of the random
effects. In both models, the random effects were estimated to be quite
close to zero, and -gllamm- essentially reports those as zeroes, while
-xtlogit- falls somewhat shy of that. That's a strange result; you
would expect that in panel data settings you would see some panel
effects, but in your case it appears to be well explained by the
demographic variables. Of course another explanation for almost any
strange result is that your model is badly misspecified, or that
random effect variances somehow fail to be identifiable. The
likelihood ratio test reported by -xtlogit- is significant, but that's
the effect of a large sample size, large enough to make the variance
of less than 1% significant.
Is that DHS data, by the way?
--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: Please do not reply to my Gmail address as I don't check
it regularly.
*
* 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/
