Date:         Wed, 17 May 2000 16:38:09 GMT
Reply-To:     Dale McLerran <dmclerra@FHCRC.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Dale McLerran <dmclerra@FHCRC.ORG>
Organization: Fred Hutchinson Cancer Research Center
Subject:      Re: Repeated measures ANOVA

Se Kim,

One respondent suggested using PROC GENMOD to fit a GEE.  [I knew I
could count on Jay to suggest GEE! :)]  There are a couple of other
options which I believe you should know about.  One is the GLIMMIX
macro which has shipped with SAS for quite a few versions now.  The
second alternative is the procedure NLMIXED, which was beta with
version 7.  It should be part of the STAT package in version 8.
I believe that each of these is better than GEE.  But each is best
suited to different types of problems.  The GLIMMIX macro is best
when covariances between errors are small (approach zero).  Typical
situations where error covariances are small are group randomized
trials where we must assume that the responses for any one member
of the group will be informative for the group as a whole.  The
NLMIXED procedure would not work well when error covariances
approach zero, but will be much better than GLIMMIX when the error
covariances are large.  In a repeated measures study, my guess would
be that the error covariances would be large, so that you would be
best off using NLMIXED.

Dale

--------------------------------------
Dale McLerran
Fred Hutchinson Cancer Research Center
Seattle, WA 98109
mailto:dmclerra@fhcrc.org
ph:  (206) 667-2926
fax: (206) 667-5977
--------------------------------------

uclaneuro@YAHOO.COM (Se Kim) wrote in
<20000516145119.4609.qmail@web3205.mail.yahoo.com>:

>Dear All,
>
>I have a question regarding an experiemnt with
>repeated measures.
>
>Patients were assigned randomly into two groups,
>treatment group and the control group. Both groups
>were followed for average of 12 months. At 3, 6, 9,
>and 12 month, four measurements were taken for each
>patient in both groups; three categorical variables
>(positive/negative test) and one continuous variable.
>
>Does Repeated measures ANOVA applied for this
>experiment or other statistical methods are more
>appropriate? Usual hypothesis in repeated measures
>ANOVA is equality of means incorporating time profile.
>But in this case, we have equality of proportions
>(three categorical measures; hbeag=hbeantigen(+/-),
>hbeab=hbeantibody(+/-),and hbvdna=hbvDNA(+/-), and one
>continuous measures; ALT level) repeated at 3 mos
>interval. The objective of this test is to compare two
>groups (e.g., we hope to prove treated group is
>better, meaning negative hbeag) at 3, 6, 9, and 12
>month.
>
>
>Any comments or help is greately appreciated.