Date:         Fri, 29 Jul 2005 17:51:11 -0700
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: Logistic Regression with a Random Effect
In-Reply-To:  <MC4-F17K4vl3brNHhmv0000d868@mc4-f17.hotmail.com>
Content-Type: text/plain; format=flowed

hein0106@UMN.EDU wrote:
>I have some binary data on cattle stillbirths (0 or 1) and I am looking to
>do a logistic regression with the data.  Some cows have more than one
>calving so I would like to put cow in my model as a random effect.
>
>Will Proc Logistic accomodate a random effect or would I have to use
>something like Proc Glimmix to get the job done?

First, PROC LOGISTIC won't accomodate that random effect you want.

Second, if your data come from a survey sample with known design
effects, you ought to be looking at using PROC SURVEYLOGISTIC,
which could accomodate the 'cow' effect through clustering in the
survey design.

Third, if you don't have a survey sample but just observational data,
then PROC GLIMMIX might be the right choice.  Or not.  It depends
on your data.

How many cows (your independent subjects I assume) do you have?
If you have a large sample of cows, then you could do this using
GEE (Generalized Estimating Equations) through PROC GENMOD.  For
a small number of cows, this is not a good choice.

Are you likely to have strong within-subject correlations?  Then PROC
GLIMMIX may not be a good choice either.  (Dale will corect me if
I'm wrong on this, but I seem to recall this applied to the original
%GLIMMIX macro at least.)

That may leave you with PROC NLMIXED to model your process.

HTH,
David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today - it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/