Usually, these are fit by using coxph. You create a survival variable
where time=1 if the product was chosen and time=2 and is censored for
the product not chosen (yet). Supposedly, the Breslow method is
equivalent to the MNL model, but I have seen the "exact" method also
used for these models.
Mike
*******************************************************************************
Michael Conklin
VP Analytic Services
Custom Research Inc.
mconklin@customresearch.com
>>> "Olivier Janssens" <OJanssens@csi.com> 01/19 5:14 PM >>>
Dear S+ users,
I am working with conditional logit model. (Model formulated to explore
by
example the consumers choice behaviour) In such model the dependent
variable
in the logistic regression can represent a variety of choice. The
potential
explanatory variables vary across the customers and for each customer
we
have one or more value for each alternative choice.
The formulation of the logit model look like the "multinomial logit
model"
but the parameter vector is assumed to be constant across categories.
Is it possible to formulate such a model in S-plus, in particular with
"multinom" designed by Venables & Ripley ...
Thanks
Olivier Janssens
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To
unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
|