Hi I have 2 groups ( Patient and Control ) and 1 psychological score. (1) I have already performed individual within group analysis. So... GLM design 1: EV1=psychological score in patients ( and of course I only run that design in patients ) contrast 1: [1,0] contrast 2:[-1,0] GLM design 2: EV1=psychological score in controls ( and of course I only run that design in controls ) contrast 1: [1,0] contrast 2:[-1,0] And then I used randomise "-D". I got some blobs. And I think I can simply compare the results between patients and controls. (2) However, I saw a paper using different approach, so I wonder whether I should also try their method? In their paper, they tried to find out brain regions where there is a significant difference in the "correlation of brain and score" in patients and control subjects. EV1=control ( so all control get 1 , patients get 0 ) EV2=patient. EV3=age. EV4=gender. EV5=total gray matter volume. EV6=score for control ( so patients get 0) EV7=score for patients.( so controls get 0) contrast=[0.0.0.0.0.1.-1] and [0.0.0.0.0.-1.1] As for the problem of demeaning: (1) maybe I can not use the additional EV8=1=intercept, because it will be very similar to EV4gender. (2) maybe I should also not use the "D" function in randomise, because it will demean every EVs, including EV1 and EV2 (3) I should just demean EV3age, EV4gender, EV5total gray matter volume. And I will keep the original scores of EV1, EV2, EV6, and EV7. Is that correct? And do I really need to repeat their method again? Or can I just use my original result? Thanks so much!!!