Date:         Fri, 10 Dec 2004 08:46:02 -0500
Reply-To:     Peter Flom <flom@NDRI.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <flom@NDRI.ORG>
Subject:      Re: Polytomous response logistic regression
Content-Type: text/plain; charset=US-ASCII

Diana

You already got some excellent advice from Dale.  (Dale always give
excellent advice).

I would only add the following:
1) Yesterday, I gave a talk on polytomous logistic regression to the NY
Area SAS Users Group.  I need to fix a couple things, but would be glad
to send it to you when that is done, if you like

2)  Regarding missing data, Dale is entirely correct that it's complex.
 But, in conversation with Joe Schafer (one of the leading experts on
missing data) he indicated that some violations of the MAR assumption
may not be bad, and that, in most cases, even when the data are not MAR,
MI is the best approach.   I don't have details on this, I believe he is
working on a book or series of articles which will discuss it, and he is
scheduled to come speak here in February......Very interesting stuff.

3) You don't say what your sample size is, or exactly how many IVs you
have, or how many levels they have.  But you have to be careful about
'slicing the pie too thin'.  If some of the combinations of levels of
IVs are rare, this will cause problems.

finally

4)  It sounds like you only want to compare certain levles of the DV.
Why?  I would first try an ordinal logistic model.


HTH

Peter

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)