[R] qualitative response model

[(Ted Harding)]

On 14-Jul-03 Z Eric wrote:
> Hi, I want to know is there other functions in R to
> estimate qualitative response model besides multinom()
> in library nnet, if this is the only possibility, I
> have a question about the application:
> for example:
> there is three transportation choice : car, bus, subway.
> each alternative has own characteristic variables,
> I want to apply conditional logit model to analysis
> the choice of three alternatives, but if some cases
> face different choice set, i.e only car and bus are
> available, how to use this function?
> I am a newcomer in R, hope the simple question not to
> bother you.
> thanks for the help.

A totally naive idea would be to incorporate the "choice"
set as a set of three factors CAR=0/1, BUS=0/1, SUB=0/1
according to availability. Naively, since (e.g.) CAR=0
would lead to zero cases observed, one might expect that
the estimation would predict zero for these. However, in
conjunction with the logistic model the parameter estimates
would be improper (-> inf).

Nevertheless, there may be some mileage in an approach which
incorporated this idea, though I'm not seeing how!

If it could be got to work, though, it should lead to a convenient
representation of interesting interactions, e.g. how does propensity
to use "car" change if "subway" is not available?

Ted.


--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at nessie.mcc.ac.uk>
Fax-to-email: +44 (0)870 167 1972
Date: 14-Jul-03                                       Time: 14:14:22
------------------------------ XFMail ------------------------------