Karl Friston wrote:
>
> Dear Esa,
>
> > We have a FDG study with two groups (athletes and controls) and three
> > levels of exertion. The exertion levels are measured at each person, so
> > this is a repeated measures experiment. We want to find out the
> > differences between the groups and the exertion levels.
> >
> > We succeeded in testing the group differences within exertion levels and
> > the effects of exertion levels.
> >
> > Our problem is that we can't test the overall group effect properly.
> >
> > How to test the group effect with SPM99? What kind of design (model)
> > and what kind of contrast? We tried in SPM Multi-group : conditions
> > and covariates. Then we defined the contrast. We tested the 6 first
> > parameters of the design matrix with a contrast 1 1 1 -1 -1 -1 , but
> > SPM stated that it is invalid contrast.
>
> This is because you were trying to use within subject variability to
> make an inference about between group effects. The simplest approach
> to the main effect of group is to average all the scans from one
> subject into a single observation and then perform a simple group
> comparison.
>
> I hope this helps - Karl
Thanks for the response. Actually we already did the average scans
of each subject, but we thought there would be a way to perform group
comparison with original scans.
Afterwards it is actually obvious that we can't do such a contrast.
Best wishes
Esa
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|