Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: Conditional multinomial logistic regression again


From   jverkuilen <jverkuilen@gc.cuny.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: Conditional multinomial logistic regression again
Date   Fri, 14 Nov 2008 12:11:28 -0500

Yes, there are at least three options in SAS:

(1) proc mdc found in SAS/ETS. There might be other SAS/ETS procs that do it but mdc fits all manner of multinomial choice models. 

(2) There is a way to trick the Cox PH program to do it by setting up the data in "super long" format akin to the way Poisson regression can be used to fit a discrete time survival model. It is just a guess but I suspect you'll find an example in the well-known Agresti 2002 categorical book. I know there is an example on Carolyn Anderson's (UIUC Ed Psych/Stat/Psych) web page under her ategorical data class. (Am on handheld so I can't give better references but "the" Google should turn up what you need.)

(3) You can also use proc nlp (in SAS/OR)  to code up the likelihood function. 



-----Original Message-----
From: "Elizabeth Allred" <lizard@hsph.harvard.edu>
To: statalist@hsphsun2.harvard.edu
Sent: 11/14/2008 11:22 AM
Subject: st: Conditional multinomial logistic regression again

Having gotten no responses to my earlier query (below), I went over to the dark side (SAS) to see what it could/would do. Using proc logistic, I can fit a conditional <binary> logistic (like clogit) and a multinomial logistic (like mlogit) but SAS, too, will not fit a conditional multinomial logistic. SAS says:

NOTE: Conditional logistic regression with polytomous response data is not
      currently supported.

Will it ever be? I'm not a programming whiz..am I asking for something impossible? Is there no way to do conditional multinomial logistic regression??

Liz

>>> On 11/13/2008 at 3:42 PM, in message <491C4AAA.6A77.00DF.0@hsph.harvardedu>,
Elizabeth Allred <lizard@hsph.harvard.edu> wrote:
> I've used clogit (conditional [fixed-effects] logistic regression but it only 
> deals with a binary dependent variable. Now I have a dependent variable with 
> 3 levels. Mlogit (multinomial logistic regression) handles that but doesn't 
> it doesn't do for the fixed effects. How can I do condtional multinomial 
> logistic??
> 
> Liz
> 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index