-------- Original Message --------
Subject: RE: nested MANOVA/ MANCOVA using SAS proc Mixed
Date: Thu, 11 Dec 2008 09:48:39 -0800 (PST)
From: Mark Belk <[email protected]>
To: [email protected] <[email protected]>
References: <[email protected]>
Hi Heidi,
You should contact Bruce Schaalje in the Statistics Department at
Brigham Young University. I have worked with him on a number of mixed
model analyses of multivariate shape data. He is an authority on Proc
Mixed, has coauthored a book on Linear Models in Statistics which
includes a chapter on linear mixed models, and he has a great deal of
experience in SAS, R, and ASREML. We have a couple of publications in
the works which will illustrate the application of mixed model designs
to heritability of shape experiments, and we will hopefully have them
out soon. I would send copies of SAS code, but I am the biologist half
of the collaboration, and it would be better to get the information
straight from Bruce.
Good luck,
Mark
Mark C. Belk, editor
Western North American Naturalist
Professor of Biology
Brigham Young University
801-422-4154
-----Original Message-----
From: morphmet [mailto:[email protected]]
Sent: Wednesday, December 10, 2008 12:20 PM
To: morphmet
Subject: nested MANOVA/ MANCOVA using SAS proc Mixed
-------- Original Message --------
Subject: nested MANOVA/ MANCOVA using SAS proc Mixed
Date: Wed, 10 Dec 2008 11:09:52 -0800 (PST)
From: Heidi Schutz <[email protected]>
To: [email protected]
Everyone,
I am wondering if anyone has used SAS Proc Mixed to run a MANOVA/MANCOVA
with RWs/PWs as dependent variables and if you have any comments that
syntax.
We need to run a 2X2 nested analysis with a random effect. GLM has
multivariate capability, but GLM always models all effects as fixed even
when the random statement is used. We have been able to figure out that
in order to run the MANOVA/MANCOVA in proc mixed we need to "trick" it
into thinking that the data are in repeated measures format, but we have
been unable to confirm whether this is the correct approach.
Here is our experimental design:
Mice are separated into two groups:
1. Control (C) 2 Selected (S)----for high wheel running behavior
(currently on generation 56)
2. Each of the above groups has 4 lines so that there are 4 S lines and
4 C lines.
3. Within each line there are also two treatment groups: 1 access to
wheels, no access to wheels.
4. Line is a random effect because there are linetype (C vs S)
differences, but there are also line differences within C and S.
Anyone out there use SAS proc mixed this way?
Heidi Schutz, Ph.D.
Postdoctoral Fellow
Biology Department
102 University Lab Building
University of California Riverside
[email protected]
951-827-2610
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
--
Replies will be sent to the list.
For more information visit http://www.morphometrics.org
