## Repeated measures binary logistc regression in SPSS

### Repeated measures binary logistc regression in SPSS

Hello all,
I would like to get in touch with someone who can help me set up a
repeated measures binary logisti regression in SPSS.
I would acknowledge your help in all the published papers.
All the best,
Deni Porej
Ohio State University
--
http://web2news.com/?comp.soft-sys.stat.spss

Hi all,

My background is in exercise science not statistics so please forgive me
for my lack of understanding in what is probably a simple query.

We are interested comparing a measure of muscle activation (DV) from 3
different muscles (IV) during a single task. In order to minimise
measurement error we have collected 3 seperate measures of the DV and have
collected data from 18 subjects with no missing values. We are interested
in the within subject effects, not the between subject effects. My initial
thought was to use a mixed analysis design (PROC MIXED) with muscle as the
repeated measure and subject as a random effect. My input data file is set
up as follows:

Sub    muscle   trial      DV
1        1        1        num
1        1        2        num
1        1        3        num
1        2        1        num
1        2        2        num
1        2        3        num

...etc

As we are not interested in between subject effects I have not included sub
as in the CLASS or MODEL statements

Data = data1;
Class sub muscle;
Model DV = muscle;
Repeated / hlps hlm type = hf subject = sub;
run;

The code works and the output from the analysis is what I would have
expected based on plotting the means and standard errors. I have searched
the literature for a similar example, but all the repeated measures designs
that I have been able to find have used a time related variable as the
repeated measure. As our repeated measure (muscle) is a within-subject
factor I believe that a repeated measures design is appropriate. My
question: Is this technique valid for our data set? The other approach that
we have considered is an ANCOVA where subject is included as a covariate,
but based on my limited knowledge, I believe this would be a less valid
technique.

Any comments would be greatly appreciated and I again apologise for the
simplicity (and likely miswording) of my query.

Regards,
Pete

2. DECENCY