st: RE: Problems with xtlogit during a simulation

["Feiveson, Alan H. (JSC-SK311)" <alan.h.feiveson@nasa.gov>]

You need to use -cap- before the -xtlogit- command to keep going after
non-convergence. Also put an iteration limit using the -iter- option .
Then check that e(converged)=1 and that _rc is zero. If both of these
are OK, then your output is valid. If either you get failure (_rc >0) or
nonconvergence (e(converged)==0), then your simulaton will go on, but
you shouldn't include the bad estimation results in your analysis.

al Feiveson




-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Joseph
McDonnell
Sent: Tuesday, August 01, 2006 1:10 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: Problems with xtlogit during a simulation

Hi

I'm looking at xtlogit in simulated samples. Each sample has 100 cases
and each case may undergo a maximum of 3 cycles. Each case has a random
age (between 25 and 40) and a constant (per-case), per-cycle chance of
success. 
The probability is based on a known, linear function of age and is
logistic in form. No other covariates are used. The simulation involves
generating a uniform random variable and comparing it to the chance of
success. If it is lower than this chance, success is 'achieved'. If
success is achieved in one cycle, no further cycles are undertaken.
Cases are not censored and 'leave' 
only after 3 cycles or the achievement of success.

However, I encounter the following message...

discontinuous region encountered
cannot compute an improvement

I'm surprised by this message. I have an extremely simple data structure
and would expect no problems in the likelihood function. But I have this
problem and it occurs quite early in the simulation suggesting it occurs
quite frequently. I've tried several options (intpoints, difficult) but
to no real avail.

I've used xtlogit many a time on real data and never had this problem.
Is the simplicity of my data set itself a problem?

I'd be interested to hear if others have had similar problems and how
they solved them.

Regards

JMcD

_________________________________________________________________
Play online games with your friends with MSN Messenger
http://www.join.msn.com/messenger/overview

*
*   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/

*
*   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/