Date:         Fri, 19 Sep 2003 11:36:29 -0400
Reply-To:     Peter Flom <flom@NDRI.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <flom@NDRI.ORG>
Subject:      Re: Multinomial Logistic Regression?  Question
Comments: To: napu1975@HOTMAIL.COM
Content-Type: text/plain; charset=US-ASCII

One question is whether you regard drug use as nominal or ordinal.  We've done work with a similar DV, regarding it as ordinal (none, Marijuana, Cocaine-Heroin, Crack, Injected drugs), but other lists of drugs may not lend themselves to such ordering.

PROC LOGISTIC can model the ordinal logistic regression, but not the nominal (AFAIK).  To do the latter, you would need to use CATMOD.

HTH



Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
www.peterflom.com
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)



>>> "Meg A." <napu1975@HOTMAIL.COM> 09/19/03 11:16AM >>>
Hello!  I am interested in developing a propensity score for  each
person in my sample.  I am planning on doing so by predicting drug use
(multi-level categorical variable) with several independent variables
to assess the comorbidity burden of the patients.  So in the model, I
will have maybe 7 independent vars predicting the multi-level drug
use.  From that I would like to find the predicted probablity (using
the beta weights from the above model).  I am familiar with the way to
do this using GLM, but since my dependent var is multinomial, I am
assuming logistic regression.  Can anyone tell me if this is correct,
or would something like catmod me more appropriate.  Also I have been
looking how to output the predicted values, but cannot seem to nail it
down?  Any help would be greatly appreciated!!

Thanks in advance!
Meg