I don't think you'll find much. There's really not much market for
multinomial probit, because it's computationally infeasible for more than
three outcomes, and hard even to do three. The reason is that its
likelihood is a (number of outcomes-1)-variate normal. It's nice in that it
doesn't require the IIA assumption of multinomial logit, so perhaps someone
with a three-outcome problem who REALLY worried about IIA (and didn't want
to use nested logit) has written something. Actually, if you have the
little Stata book on ML programming, you could probably do a quick
three-outcome model by cribbing their bivariate probit example, replacing
the normal stuff with bivariate normal. It's not hard to write down, just
hard for the computer to calculate. You might also see Daganzo's book on
probit for more on why not to do it. EJ
-------------------------
Eric R. Jensen
Professor of Economics
Box 8795, College of William and Mary
Williamsburg, VA 23187-8795
(757) 221-2384 Fax (757) 221-1175
home page: faculty.wm.edu/erjens
------------------------------
Date: Mon, 10 Jun 2002 10:34:43 +0700
From: "Daniel Muller" <[email protected]>
Subject: st: RE: multinomial probit
I asked the same question a while ago and was only pointed to a multiVARIATE
Probit (-findit triprobit-). I still didn't find an ado for the multiNOMIAL
case and would also appreciate a reference to one.
Greetings,
Daniel
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]]On Behalf Of M. Nizam Khan
> Sent: Saturday, June 08, 2002 11:37 PM
> To: [email protected]
> Subject: st: multinomial probit
>
>
> Dear statalisters,
>
> Does stata have any program to estimate multinomial probit?
>
>
> Thanks
>
>
> Nizam Khan
*
* 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/
