Re: st: RE: re: nlogitrum

["R.E. De Hoyos" <redeho@hotmail.com>]

Flora,

When -nlogit- fails to be consistent with the IIA is because the categories 
within each branch have a parameter that is not stable to additions (or 
eliminations) of one of the choices. Within a utility maximizing framework, 
one can say that the degree of substitution (or analogously, 
complementarity) between any two choices is the same (this is the same 
for -clogit- or -mlogit-).

What -clogit- is doing is imposing the IIA restriction only for choices 
within each branch. You can try a different partition of your branches to 
see if you can get around the IIA.

Hope this helps,

Rafa
________________________
R.E. De Hoyos
Faculty of Economics
University of Cambridge
CB3 9DE, UK
www.econ.cam.ac.uk/phd/red29/

----- Original Message ----- 
From: "Flora Au" <fau@ucalgary.ca>
To: <statalist@hsphsun2.harvard.edu>
Sent: Thursday, May 05, 2005 4:33 PM
Subject: Re: st: RE: re: nlogitrum



> Thank you Nick. To make the issue that I am facing a bit more clear, I
> first started with a conditional logit model and tested for IIA (which I
> also used the old fashion hausman command to double check). As a result,
> my model failed the IIA, then I test my model with a nested logistic model
> 'nlogit'. However, I got the same result as my 'clogit' model, so my
> colleague suggested to used the 'nlogitrum' instead. By what I describe
> above, could anyone tell me why I have the same results using -clogit- and
> -nlogit- ?
>
> Thank you again,
> Flora
>
> Nick Cox said:
> > This was previously posted on 28 April
> > and got no replies.
> >
> > Why not? I guess because only a few
> > kinds of rather basic comment seem possible.
> >
> > First, it is possible that your data
> > are not suitable for this model.
> > You need to think about that.
> >
> > Second, -nlogitrum- offers handles for
> > varying the maximization process in using
> > -ml- to fit the model, particularly important
> > when defaults run into problems. Those are documented
> > and explained in various standard places, e.g.
> > documentation on -ml- and the Gould and
> > friends book on the same area of Stata.
> >
> > Nick
> > n.j.cox@durham.ac.uk
> >
> > Flora Au
> >
> > > I am having problem with NLOGITRUM.
> > >
> > > This the the command I am using to run the model:
> > >
> > > nlogitgen branch = alt(package:1|2, nopack:3)
> > > nlogittree alt branch
> > > nlogitrum pref nurse nsq na2 na3 n_nt1st n_highedu homevis
> > > visitsq va2 va3
> > >  hv_nt1st hv_highedu nhv homehelp helpsq cost c_highinc, group(indid)
> > >  nests(alt branch)
> > >
> > > The first two command ran fine, however the last command result in an
> > >  error message which is written as follow:
> > >
> > > could not calculate numerical derivatives
> > > flat or discontinuous region encountered
> > >
> > > Can you assist me as to how to fix this model?
> > *
> > *   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/
>
>
> *
> *   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/
*
*   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/