Date:         Mon, 29 Sep 2008 15:12:29 -0500
Reply-To:     Robin R High <rhigh@UNMC.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Robin R High <rhigh@UNMC.EDU>
Subject:      Re: Proc genmod grouping levels
Comments: To: cyc1120 <cyc1120@GMAIL.COM>
In-Reply-To:  <6c1a4a17-74c1-4ba0-a0ec-e0aae15d1d05@m36g2000hse.googlegroups.com>
Content-Type: text/plain; charset="US-ASCII"

It's possible with a format for a class variable, e.g., having group coded
as 1,2,3 and and then placing groups 1 and 3 into one level to compare
with group 2:

PROC FORMAT;
VALUE gr 2='2' 1,3='1,3';
run;

PROC GENMOD DATA=ids ;
CLASS group cond ;
MODEL respns/trial = group|cond / dist=binomial link=logit type3;
FORMAT group gr.;
RUN;

Robin High
UNMC




cyc1120 <cyc1120@GMAIL.COM>
Sent by: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
09/29/2008 11:14 AM
Please respond to
cyc1120 <cyc1120@GMAIL.COM>


To
SAS-L@LISTSERV.UGA.EDU
cc

Subject
Proc genmod grouping levels






Hi,

I was wondering if there is an easy way for proc genmod to group
levels for the covariates?

For example, let's say I have 5 levels for the Area variable (A1, A2,
A3, A4 and A5). I would like to group A3 and A5 together. Is there any
other ways other than do it manually from the initial dataset?


Thanks