Date: Wed, 22 Nov 2006 21:01:55 -0500
Reply-To: S Elgie <firstname.lastname@example.org>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: S Elgie <email@example.com>
Subject: Re: More on Repeated Measures
Content-Type: text/plain; format=flowed
The design I am testing has eight measures at three time periods, and I am
conducting a multivariate repeated meassures analysis. The design is
completely repeated with no between-subjects factor(s). I do get
univariate results, but what is puzzling me is the two tables containing
multivariate tests (Pillais and so on), one labelled 'Multivariate Tests'
and the other 'Tests of Within-Subjects Effects.'
>From: "Swank, Paul R" <Paul.R.Swank@uth.tmc.edu>
>Reply-To: "Swank, Paul R" <Paul.R.Swank@uth.tmc.edu>
>Subject: Re: More on Repeated Measures
>Date: Wed, 22 Nov 2006 16:00:56 -0600
>The multivariate tests assume multiariate normality and homegenous
>variance covariance matrices across groups, the univariate repeated
>measures assumes homegeneity of variance covariance matrices, sphericity
>of the variance covariance matrix and normality of the residuals. The
>univariate repeated measures is typically mor sensisitve since the
>violation of sphericity tends to result in more significant results. The
>tests are different and do have different degress of freedom.
>Paul R. Swank, Ph.D.
>Professor, Developmental Pediatrics
>Director of Research,
>University of Texas Health Science Center at Houston
>From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
>Sent: Wednesday, November 22, 2006 2:39 PM
>Subject: More on Repeated Measures
>Thanks so much for all help received on repeated measures. I am
>emboldened to ask a further question. When one does a multivariate
>repeated measures analysis, there are two tables of multivariate
>results. The first is labelled 'Multivariate Tests' and contains tests
>of the intercept and within subject factor. (For now let's assume just
>one repeated factor). The second table is labelled 'Tests of
>Within-Subjects Effects' and contains just the tests of the repeated
>factor. I have not ever had the problem of these two tables giving
>substantially different answers, such as one significant and the other
>not. But the numbers and degrees of freedom etc.
>are different and I have never been sure why and particularly not sure
>which one should be reported.