--- Stephan Brunow wrote:
> how can it happen that a mlogit converges properly and
> mprobit reportsthat the likelihood-value is not concave?
> Is there some literature clarifying that issue?
The two models are pretty similar. If you look inside the
-mprobit- code you will see that -mprobit- uses -mlogit- to
create starting values. However, -mlogit- is such a common
model that StataCorp has taken the extra effort to write
special code for it, while -mprobit- uses Stata's d1
maximization program. If your model fits well, both should
work, but if it doesn't than -mlogit- is more likely to find
a maximum than -mprobit-. A common problem occurs if one of
the categories in your dependent variable is pretty rare.
The solution is to combine that rare category with some
other categories. See the example below:
*------------ begin example ----------------
set more off
sysuse auto, clear
mlogit rep78 price mpg foreign
/*stops after 100 iterations*/
mprobit rep78 price mpg foreign, iter(100)
recode rep78 1/2 = 3
mlogit rep78 price mpg foreign
mprobit rep78 price mpg foreign
*-------------- end example ------------------
(For more on how to use examples I sent to the Statalist, see:
http://home.fsw.vu.nl/m.buis/stata/exampleFAQ.html )
Hope this helps,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
*
* 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/
