Dear Esa,
> Our data is as follows: 11 subjects and 8 scans per subject, which
> makes 88 scans. At each scan, each subject gave a subjective rating (0
> -10).
>
> We are trying to do a correlation analysis between subjective ratings
> and actual voxel values. The problem is that our data is not
> independent i.e. does not follow the assumption for correlation
> (regression analysis).
>
> Is it possible to analyze this kind of data using correlation
> analysis? Anyway, we tried the following model:
>
> Multi-subject : Covariates only, Covariate interaction with
> subject. Is this correct?
Yes - this models the main effect of the covariate and the covariate x
subject interactions (i.e. computes a regression coefficient for each
subject). You can test for the significance of the average regression
(or indeed differences among the subjects) with the approproate
contrast.
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|