Date:         Tue, 25 Jun 2002 11:46:27 -0400
Reply-To:     melanie@gtw.net
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Melanie Rogers <melanie@gtw.net>
Subject:      Hotelling Tsquared
Content-Type: text/plain

I'm new to the list, and have an spss question that I'm hoping someone out there can help me with.

I have a data set that just has one group (students).  Each student completed 17 dependent measures (measures correlate with each other anywhere from .33 to .86) twice.  Because I have a within-subjects design with multiple DVs (and just one group) my reading of several stats books tells me that Hotelling's Tsquared  is the best way to deal with my data.  How do I do this?

So far, I have run a MANOVA with the following syntax:

manova math1 math2 hx1 hx2.....
 /wsf=time(2).

Results show me a significant (p = .008) Hotellings Trace value.  I'm stumped as to what I should do next, however.  Do I now run separate Hotelling tsqauared tests on each of the pairs of dependent measures (e.g. math1 and math2) and look at the reliability between scores as the syntax below would give me?

reliability var=math1 math2
 /statistics=hotelling.

I know that hotelling tsquare tests are the multivariate equivalent of t-tests, so it seems wrong to me that I should now run separate hotelling tsquare tests for each of my pairs of dependent measures.  However, I am not sure how else to see on which dependent measures students significantly differ between Time 1 and Time 2.

Thanks for any advice you can offer.