Date:         Tue, 7 Jun 2005 15:35:23 -0700
Reply-To:     cassell.david@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <cassell.david@EPAMAIL.EPA.GOV>
Subject:      Re: multinomial logistic regression - with slight problems
In-Reply-To:  <c2a8be6d05060713016944242d@mail.gmail.com>
Content-type: text/plain; charset=US-ASCII

Kath Santos <flowerpotcat@GMAIL.COM> replied:
> A related question regarding your suggestions... Does Proc
> SurveyLogistic handle multinomial logistic regression for complex
> samples? I've tried to do it, SAS did give an output but I didn't know
> how to interpret it (it looks so much different from the catmod
> output). It seems it didn't take into consideration the different
> levels of the dependent variable... Since I don't know how to do
> multinomial logistic in the proc surveylogistic, i did my runs in
> stata instead.... but if any of you could let us know how to do it in
> sas i'd appreciate it so much.

[1] PROC SURVEYLOGISTIC does indeed handle multinomial logistic
regressions.  Just specify the link function as GLOGIT.

[2] The output for generalized logits regression in PROC SURVEYLOGISTIC
is actually designed to look a lot like a similar analysis in PROC
LOGISTIC, rather than that in PROC CATMOD.

[3] If it works better for you in Stata, then go for it.  You don't
have to stick with SAS if you're fluent in multiple programs.

[4] The easiest way to get a handle on the output of PROC SURVEYLOGISTIC
is to read the SAS Online Docs.  Go to sas.com and track down the
online documentation for SAS 9.1 .  Then read over the examples for
PROC SURVEYLOGISTIC in the SAS/STAT section, and you'll see.  You
can also jump over to the examples for PROC LOGISTIC and see how
LINK=GLOGIT looks in that proc.

HTH,
David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician