home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky bit.listserv.sas-l:5507 bit.listserv.stat-l:2297
- Path: sparky!uunet!dziuxsolim.rutgers.edu!gandalf.rutgers.edu!parilis
- From: parilis@gandalf.rutgers.edu (Gary Parilis)
- Newsgroups: bit.listserv.sas-l,bit.listserv.stat-l
- Subject: multivariate design question
- Message-ID: <Jan.4.22.55.03.1993.29092@gandalf.rutgers.edu>
- Date: 5 Jan 93 03:55:04 GMT
- Followup-To: bit.listserv.sas-l
- Organization: Rutgers Univ., New Brunswick, N.J.
- Lines: 45
-
-
- I'm no multivariate statistics wizard, and I'm having a bit of trouble
- figuring out how to set up the analyses for a set of data that looks
- like this:
-
- I'm looking for differences between 2 levels of a discreet independent
- variable (treatment A and treatment B) on several related continuous
- dependent variables.
-
- There are 2 groups (group1 and group2). Data was collected at three
- times (t1, t2, t3), in a repeated analysis design.
-
- For both groups, t1 is the baseline.
-
- For logistical reasons, group1 got treatment A at t2 and treatment B
- at t3, and group 2 got the treatments in the reverse order.
-
- The ideal situation would have been a simple between Ss design, in
- which one group was only given one treatment, and the other group was
- only given the other, but unfortunately, because of manipulation
- involves different treatments of students (which are hypothesized to
- differentially affect their academic performance) it would be
- unethical to do so.
-
- I'm using SAS. A univariate version of what I want to do would look
- something like:
-
-
- proc glm;
- class group;
- model y1-y3 = group / nouni;
- repeated time 3;
-
-
- How do I set up a multivariate version of this?
-
- I don't want to resort to doing separate univariate analyses, for
- obvious reasons...
-
-
- Thanks muchly in advance...
-
- -Gary
-
- parilis@gandalf.rutgers.edu
-