Re: SPSS for repeated measures on a dichotomous variable

Sarah Berger wrote:
I am trying to analyze a dichotmous variable in a repeated measures
design with 1 between-subjects factor. A quick search through the
archives has turned up several suggestions to use GEE for this
analysis. To make things a bit more interesting, it is a complex
comparison, where the scores are weighted so that 3 trials are compared
to a 4th. This has not been a problem when my variables are continuous
because I just run a repeated measures ANOVA with a difference
contrast.
Is GEE still appropriate in the case of a difference contrast? Can I
run GEE in SPSS? Is there sample syntax available somewhere that I
could look at?
If I run a rep. meas. ANOVA on a dichotomous variable is there
something I can adjust to account for the wrong type of data?
Thank you for your help,
Sarah

With using dichomotous data I think you just need to check the
assumptions of the repeated measures model - normality and multi-sample
sphericity being the main ones here, I think. If the dichotomy doesn't
work too well then the ANOVA assumptions are badly violated (probably
when the mean is close to 1 or 0). Using the correct model might be
more powerful even if the ANOVA assumptions are met - I don't know
enough about GEE, but I'd probably use a multilevel logistic model as
an alternative.

Thom

.

Relevant Pages