Date:         Fri, 8 Oct 1999 08:49:50 GMT
Reply-To:     Aki Akfak <a_akfak@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Aki Akfak <a_akfak@HOTMAIL.COM>
Subject:      Re: repeated measures analysis-multivariate?
Content-Type: text/plain; format=flowed

This message was posted by A.Akfak for Michael F.

You wrote:

>I have two data sets(before treatment & after treatment, 29 subjects)
>I measured the back extensor strength of each subject at 7 specific
> >angles,0 degree, 12 degree, 24 degree, 36 degree, 48 degree, 60 >degree,
>72 degree. After 8 weeks training, I measured the back >extensor strength
>of the subject again.

>I'd like to know whether the diffence of strength is significant at >each
>specific angle after training and the improvement of strength >at each
>specific angle is different

>What should I do?(I have both the SPSS and SAS.)

The number of subjects is not much greater than the number of observations
per subject which suggests paying some attention to the correlation
structure.  However,a multivariate response model would be vastly
overparameterised.  On the other hand,a repeated measures model may be too
simple.

Whether there is any merit in looking for an intermediate approach could
depend upon: what background information you have regarding the effect of
the different angles, whether you have adequate sample size to reliably
estimate the treatment effect, and the correlation patterns evident in your
data.  Exploratory plots will help with the latter, and may suggest some
simplification.  Summarising the response profiles over different angles in
some convenient way, eg by their linear trend, could enable a very simple
analysis, and may be as much as your data can support.

Michael F.


______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com