Thank you Paul for your suggestion and the very useful papers!
I tried this model by using a Full factorial design with one factor with two levels (patients-controls) and modelling the volume as a covariate with interaction with factor 1. Then I used the contrast: 0 0 1 -1
I believe this model will calculate in which voxels the slope of the volume vector is steeper in patients than controls.
Is this correct?
If correct, do you have any suggestion on if, or how, the volume vector values should be centered?
Best Regards
Carl Johan Ekman
|