--- On Tue, 27/10/09, Jennifer Schmitt wrote:
> I am trying to run xtmelogit with fixed effect dummy
> variables for village, but also include a random effect (a
> random intercept) for village.
Now you have two intercepts for each village, one random and
one fixed. That is obviously not identified, so there is no
way in which Stata or any other software can give sensible
estimates of such a model. Think of it this way: The dummies
have explained away all the variance at the village level,
so what is there left for the random intercept to explain?
In this case the random variance should be 0. But what if
we started with the random intercept? Now the random
intercept will do its best to explain as much as possible,
and the dummies will pick up whatever is left. So there
are multiple sets of possible estimates and Stata has no
way of choosing between them. This is what we mean when
we say an estimate is unidentified.
The bottom line is that if you specify this model you
should not get estimates, as these estimates would be
nonsense. So it should either not converge, or produce an
error message. Having said that, the error message you
got is a bit too cryptic and there is definately room
for improvement there.
-- Maarten
--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany
http://www.maartenbuis.nl
--------------------------
Send instant messages to your online friends http://uk.messenger.yahoo.com
*
* 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/
