DMRM-SAS.txt Doubly Multivariate Repeated Measures ANOVA on SAS --------------------------------------------------------------------------- Date: Wed, 23 Nov 1994 19:57:49 GMT Sender: Stat-l Discussion List From: hang-yue ngo Organization: University of Chicago Subject: Doubly Multivariate in SAS To: Multiple recipients of list STAT-L A quick question: Does any SAS procedures do the doubly multivariariate analysis? For example, at each treatment level, 2 variables are observed for each subject over 3 time points. So for each subject, the variables are: (Y1 Z1), (Y2, Z2) and (Y3, Z3). How can I analysze the data using SAS? How about the other software, such as SPSS and BMDP? --------------------------------------------------------------------------- Date: Wed, 23 Nov 1994 22:10:04 GMT From: "Randall D. Tobias" Subject: Re: Doubly Multivariate in SAS To: Multiple recipients of list STAT-L GLM in Version 6.10, via the new IDENTITY repeated transformation. |> For example, at each treatment level, 2 variables are observed for |> each subject over 3 time points. So for each subject, the |> variables are: |> (Y1 Z1), (Y2, Z2) and (Y3, Z3). |> How can I analysze the data using SAS? Something like PROC GLM; CLASS TMT; MODEL Y1 Y2 Y3 Z1 Z2 Z3 = TMT; REPEATED VAR 2 IDENTITY, TIME 2 POLYNOMIAL; should do the trick. -- Randy Tobias SAS Institute Inc. sasrdt@unx.sas.com (919) 677-8000 x7933 SAS Campus Dr. 72450.2545@compuserve.com (919) 677-8123 (Fax) Cary, NC 27513-2414 --------------------------------------------------------------------------- Date: Fri, 25 Nov 1994 08:58:03 -0500 From: Robert Terry Subject: Re: Doubly Multivariate in SAS To: Multiple recipients of list STAT-L Yes, GLM will do it via the M statement. The M statement allows you to construct contrasts among the dependent variables listed on the left hand side of the equal sign of your model statement. To construct doubly multivariate contrasts, simply apply within-effect contrasts to each of the "dependent measures". A simple, elegant way of conceptualizing this is to construct your within-effect contrast matrix, C, and an identity matrix, I, with dimensions equal to the number of DV's. The Kronecker product of C x I will produce the contrast matrix to use in the M statement (but be careful in how you "order" the DV's in the model statement). Robert Terry Duke University