Date:         Fri, 21 May 2004 16:28:49 -0500
Reply-To:     Paul R Swank <Paul.R.Swank@UTH.TMC.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Paul R Swank <Paul.R.Swank@UTH.TMC.EDU>
Subject:      Re: Paired ANOVA?
Comments: To: Michelle <micray2000@NETSCAPE.NET>
In-Reply-To:  <d45f8e97.0405211005.7b30c10b@posting.google.com>
Content-Type: text/plain; charset="us-ascii"

Proc freq will do the extension of the McNemar to more than 2 categories,
the Bowker test. It works for any square table.

Paul R. Swank, Ph.D.
Professor, Developmental Pediatrics
Medical School
UT Health Science Center at Houston


-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
Michelle
Sent: Friday, May 21, 2004 12:05 PM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Paired ANOVA?


Hi all,

I have a group of people that completed a survey both before and after an
intervention.  I'm interested in seeing how certain variables changed.  I
did a McNemar's test on the dichotomous variables, but my question is on the
ordinal variables that have 3 categories.  So, I have a variable called
forgmed (how many days did you forget to take your meds?) which was asked
before the intervention, then I have a variable called forgmed_t1 which was
asked after the intervention. The responses are 0, 1 or 2 (0 days, 1 day or
2 or more days).  Should I be using a repeated measures anova in proc glm?
If so, what is my class variable and what would my model and contrast
statements look like?  Any help would be much appreciated!

Thanks,

Michelle