Hi eugene
the FSL Web docs and several previous posts affirm that if i want to control for a covariate (but not investigate group*covariate interaction) I create one column with demeaned values (around the ungrouped mean).
This is equivalent to orthogonalize the covariate wrt the group mean , that is, when comparing the mean of two groups (with different mean values in the covariate), investigate the differences that would exist if the two groups had similar covariate values.
where am i wrong???
cheers
Alberto
|