Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Concavity problems with clogit


From   "Verkuilen, Jay" <JVerkuilen@gc.cuny.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Concavity problems with clogit
Date   Wed, 6 Feb 2008 17:08:35 -0500

Peter Dijkstra wrote:

>>And another problem I incurred with conditional logit (sorry for
sending so soon again, I had this problem a few days ago but apparently
I did not sent the email in a correct way). While using clogit combined
with xi, I obtain the error "(not concave)" after almost each iteration.
Besides, each iteration takes ages. Did I define too many variables or
are the variables too much correlated? Cheers, Peter<<

To check the correlation issue, take a look at standard co-linearity
diagnostics, e.g., the "collin" package. Exact collinearity should cause
Stata to complain but you may simply have a large ill-conditioned
problem. Beyond that, how many parameters do you have? In general,
estimation speed is proportional to the square of the number of
parameters plus the number of observations. If it's not concave, I'm not
surprised it's taking forever to iterate as the optimization algorithms
are probably working overtime to try to reduce step size, ridge, etc. Is
there a simpler model you can fit to see where things might start to go
wrong, such as one that has fewer predictors?

Jay 


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index