[Limdep Nlogit List] panel data model

[William Greene]

Phil. As you note, you can indeed just create the 11 dummy variables
and fit a fixed effects model, which is what I would do if these were
my data.  For a random effects model, your problem is this: In order
to compute the log likelihood for the random effects model, it is 
necessary to compute the product of the probabilities before taking the
logs - the log-L is not the sum of the logs; it is the log of a sum of
the products.  So, you are looking at computing the product of 2,000
probabilities.  If probabilities are on the order of .5, then you are
trying to compute .5 to the 2000 power., which will either (1) be pure
rounding error when you are done or, far more likely, (2) cause an
underflow and crash the program.  LIMDEP is very averse to (2), so it
will not let you try.  A few hundred observations per group is as far
as it will let you go.  (Oddly enough, I received the same question, 
though about a Poisson model, only a couple days ago. My answer was the
same.)
Bill Greene

----- Original Message -----
From: "Philip A. Viton" <viton.1 at osu.edu>
To: limdep at limdep.itls.usyd.edu.au
Sent: Friday, August 15, 2008 12:02:11 PM GMT -05:00 Columbia
Subject: [Limdep Nlogit List] panel data model


I'm trying to estimate a panel-data version of the binary logit model 
for a sample of votes of judges on the courts of appeals. In this 
case, my groups are the 12 circuits; within each circuit I have a 
large number ( > 2000) of individual votes. But when I try to 
estimate the model I get error 548, referring to an abnormally large 
group-size variable (which I suppose it is).

Have I run up against a  built-in limitation of limdep (v 8), or have 
I set things up wrong? (Of course, I can estimate the fixed-effects 
model by hand, by setting up dummy vars for the circuits, and not 
worrying about the panel specification).

Thanks!

------------------------
Philip A. Viton
City Planning, Ohio State University
275 West Woodruff Avenue, Columbus OH 43210
viton.1 at osu.edu

_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au