[Limdep Nlogit List] Another NLOGIT/LOGIT Question

[William Greene]

Shane. I'm familiar with both.
You cannot be fitting the same model. I suspect you should have
a dependent variable that is 0 for the nonchosen choices and 1 for the
choice, and add ;WTS=weight to the nlogit command.  Even with weights,
a clogit model will converge in a handful of iterations. Your setup
can't be right.
Bill Greene

----- Original Message -----
From: "Shane L. Martin" <slmartin at mitre.org>
To: limdep at limdep.itls.usyd.edu.au
Sent: Thursday, June 3, 2010 2:18:57 PM GMT -05:00 US/Canada Eastern
Subject: [Limdep Nlogit List] Another NLOGIT/LOGIT Question

Hi,

I was hoping that someone who is familiar with both LIMDEP/NLOGIT and STATA might be able to help me out on this:

When I run the following model in STATA, it converges in 4 iterations:
clogit choice fare itdist circuity gates concs runways direct seat [fweight=weight] , group(obs)

weight measures how many people made the exact same choice facing the exact same choice set.  Group(obs) is just an identifier to indicate which lines of data go together to make an observation.


However, using the EXACT SAME DATASET, when I run the following in NLOGIT, it fails to solve, even with 1500 iterations:
CLOGIT
;Lhs=weight,nij
;Rhs=fare, itdist, circuity, gates, concs, runways, direct, seats
;maxit=1500
$

In this case, weight is the frequency variable, nij tells how many lines of data belong to an observation - from what I can tell, a similar thing to the group(obs) in STATA.

Any ideas?  I'm totally stumped on this. Thanks,
Shane

Shane L. Martin, Ph.D.
Economist
The MITRE Corporation
Center for Advanced Aviation System Development
703-983-3902
slmartin at mitre.org

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