Statalist


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

RE: st: ml appears to converge, but won't admit it?


From   Maarten buis <[email protected]>
To   stata list <[email protected]>
Subject   RE: st: ml appears to converge, but won't admit it?
Date   Thu, 24 Jan 2008 13:47:54 +0000 (GMT)

--- Marc Pelath wrote:
> After reading through Gould et al's excellent book on maximum 
> likelihood estimation in Stata, I've been attempting to estimate some
> more and less complicated models using the d0 method.  But in
> executing ml maximize, I often (not always) run into the following
> paradox:
>
> -          The algorithm backs up until I get bored watching it,
> suggesting some kind of problem with the likelihood function.
>
> -          Stopping the search, then using ml plot shows that it has
> apparently maximized the likelihood with respect to each parameter
> individually. 
>
> -          The point at which it's stuck appears to be a plausible
> solution, economically.


We cannot give you any specific advise if you don't tell us what your
model is. So I will only give you some general comments:

I have made some silly mistakes in the past that lead to similar 
results. The two most common ones are:
1) not specifying the dependent variable in -ml model-, which can give
some surprising results, and
2) specifying the likelihood instead of the log-likelihood

Other things to think about are 
1) Numerical precision may be a problem: things like 1-normal(`xb') are
better modeled as normal(-`xb') 
2) have you used all the simplifications that result from taking the
logarithm?
3) reparamerize the model, other parameterizations may have
surprisingly big effects on how well the likelihood behaves.

The most direct advise I can give you is start with a simpler model,
make sure that that works and converges, and than add complications one
at the time till your model no longer converges, and than you have
found your problem. 

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/
-----------------------------------------


      __________________________________________________________
Sent from Yahoo! Mail - a smarter inbox http://uk.mail.yahoo.com

*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index