[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Profile likelihood

From   Maarten buis <>
Subject   Re: st: Profile likelihood
Date   Wed, 17 Dec 2008 12:33:42 +0000 (GMT)

--- Katia Bobulova <> wrote:
> Dear Stas,thank you very much for your suggestion, but it doesn' t
> work because a and d are not numerical variables. They are
> probabilities that I want to estimate with my maximum likelihood. So
> Stata gave my an error message. <snip>

My impression of this discussion is that you tried to estimate a model
using -ml- and could not get it to work, and than tried to solve that
problem by making a complicated problem more complicated (by turning to
profile likelihood rather than regular likelihood). I am reading
between the lines here, but if this is your debugging strategy, then it
is absolutely doomed to fail. The only way to debug such a problem is
to reduce to problem to an easier problem, solve that, and only once
you have solved that continue to a slighly more complicated problem,
solve that, and only once you have ... etc. I usually start with either
a logistic regression likelihood or a regression with normal errors,
and gradually continue form those models. (actually, I usually start
with the complicate model I am interested in, than I fail, and only
than do I follow my own advise and start with those simple models).

-- 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 N515

+31 20 5986715

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index