Dear Alfonso

As always, I highly appreciate your answer! THANK YOU!

Just to make sure I understand your point on the MVPA statistics. Running non-parametric stats for MVPA-components (eye(8)) on the contrast of main interest (in my case: drug>placebo, rest post>rest pre) gives me an empty brain. So, I can see two choices: to give up the MVPA method or to use parametric statistics instead, which show resuls (and select seeds with them and then run a post hoc afterwards). I lack the feeling how acceptable / correct it would be. Or maybe there are some other settings to play with? What would you do in my place?

Are the values: height threshold < 0.001 p-uncorrected and cluster threshold p < 0.05 FEW- (or FDR-)-corrected right there? I guess, the non-parametric stats with 1000 simulations should also use the same values, right?. Also, is there any smallest number of voxels for selecting seeds from MVPA that is a reasonable number for setting a seed (like 15, 20, 30, 40, 50...)?

I would be very thankful for your comments on that

Lucas