Re: st: SAS vs STATA : why is xtlogit SO slow ?

["Joseph Coveney" <jcoveney@bigplanet.com>]

k7br@gmx.fr wrote:

. . . Iteration 0:   log likelihood =    -1.#INF
Iteration 1:   log likelihood =    -1.#IND
Hessian is not negative semidefinite

. . . when I compare SAS's results with STATA on a MUCH
(really much) smaller sample (less than 2000 observations, 146
individuals, 11 points on average per individual) then the results are
exactly the same between the two systems (same point values + standars
errors+ P-values)... thus suggesting that something bad is going on
when STATA try to fit the fixed effect logit model on a larger dataset
...
So I am puzzled ...

What do you think ?
Thanks again for your help

--------------------------------------------------------------------------------

I would hate to have this thread end on the sentiment that "something bad is
going on when STATA try to fit the fixed effect logit model on a larger
dataset", but I'm afraid that unless you start providing relevant details, no
one is going to be able to help you much more.  Take a look at the list's FAQ
(URL below), especially peruse Sections 3.1 to 3.3, and come back with something
substantive that we can respond to.  (You might want to give Section 8.2 a quick
glance, too.)

Why do you assume that the problem is Stata's?  The tenor of your postings here
have been in the same vein as if you'd posted something on SAS-L to the effect
of, "Why does SAS claim convergence in a fixed-effects logistic regression model
when I have it on good authority that the log-likelihood is indeterminate and
the Hessian is not negative semidefinite?".  (By the way, does SAS actually
claim convergence?  You haven't shown us much of anything about your problem,
including anything from either the SAS .LST file or .LOG file, to substantiate
your assertions.)

Joseph Coveney


*
*   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/