Re: st: nlogit nesting scheme for 3-level degenerate model: datastructure

[John Fulton <fulton@ssc.wisc.edu>]

Thanks.  I'm not sure if this is relevant to your problem, but it's my 
impression that Heiss redid his data to agree with reference categories 
used in Greene's textbook (2003 edition).

Assuming nlogit is appropriate, how should I structure data for a 
degenerate nest in a 3-level model?


Anders Alexandersson wrote:

> Oops, both articles from Heiss are from 2002. I was also looking at
> Silberhorn et al. (2006), hence my confusion.
>
> Silberhorn, N., Boztug, Y., and Hildebrandt. Estimation with the
> nested logit model: Specifications and software peculiarities. SFB 649
> Discussion paper 2006.017.
>
> Anders Alexandersson
> andersalex@gmail.com
>
> On 1/2/07, Anders Alexandersson <andersalex@gmail.com> wrote:
> > On 12/30/06, John Fulton <fulton@ssc.wisc.edu> wrote:
> > > I'm setting up data for nlogitrum.  In terms of the nlogitrum
> > > command, I have two nests at the altsetvar1 level (top level).  One of
> > > these is degenerate with an elemental choice at the altsetvarB 
> > (bottom)
> > > level.  What do I do for the altsetvar2 level for this nest?  
> > Should it
> > > have a nest of its own that is degenerate on level 2 as well?  None of
> > > the examples of data management for degenerate nests I've found have
> > > dealt with three-level models.
> >
> > I think that you should not specify the altsetvar2 level in
> > -nlogitrum-, if you want the equivalent three-level non-normalized
> > nlogit model with degenerate nests from -nlogit-. Heiss (2006, 242)
> > mentions that the dissimilarity parameter of degenerate tests is not
> > defined in the nlogit rum model, and gives an example (on pages 245
> > and 251, i.e. model "M" = model "N").
> >
> > I found Heiss (2006) article instructive, but I think it contains some
> > errors. That is, I needed Heiss (2002) draft article to be able to
> > reproduce most of the examples in Heiss (2006). I have still not been
> > able to reproduce Heiss (2006, 245 and 251) 3-level model "N" (model
> > "L" in Heiss 2002) because the -nlogitdn- example gave me syntax error
> > 198.
> >
> > I note in the 3rd line of
> > http://www.stata-journal.com/software/sj2-3/st0017/nlogitdn.ado that
> > -nlogitdn- is based on version 7.0.7 of -nlogit- whereas the current
> > Stata 9 version of -nlogit- is 7.1.6. I have not yet tried -nlogitdn-
> > with the older version of -nlogit-. Are -nlogitdn- and the current
> > version of -nlogit- incompatible? Is someone else able to reproduce
> > model N in Heiss (2006) with a current version of Stata?
> >
> > References:
> > Heiss, F. 2006. Structural choice analysis with nested logit models.
> > The Stata Journal 2(3): 227-252.
> > Heiss, F. 2002. Specification(s) of nested logit models. Online at
> > http://www.mea.uni-mannheim.de/mea_neu/pages/files/nopage_pubs/dp16.pdf.
> >
> > Anders Alexandersson
> > andersalex@gmail.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/
*
*   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/