>> I don't like the idea of including MP only
>> for at least one reason: "Imagine you have a GM voxel between two WM
>> voxels then movement in opposite direction would lead to a signal
>> change in the same direction which you cannot model with MP only,
>> right?"
>
> Yes... It would be interesting to see where the MP2 effects appeared...
>
Just to clarify, the F-tests were all made in the full model, but on
e.g. the MP2 subspace. The compleete set (24 reggressors) is in my
experience not compleetely orthogonal, so could this be an effect of
some of the MP2 variance beeing explained by a linear combination of
the remaining 18 regressors?
Best
Torben
|