Hi Anderson
Thank you again very much for your help. I am a bit confused about the pure between effect. Sorry if I repeat somethings you already wrote. I just want to make sure I understood correctly.
Again, for example in the factorA*factorB*group design. If I want to test if patients have higher activity in factor A1 > A2 compared to controls, this would then be the factorA*group interaction. For which I would use within (and whole) permutations because it isn’t a pure between effect.
So this isn’t then the same as if I would first build the contrasts in each subject individually (A1B1+A1B2-(A2B1+A2B2)) and permute them to test group differences. Does this then test a different hypothesis or is this just not a valid way to test the hypothesis I’m interested in? I also tried both ways and they indeed give very different results (with within permutations highly significant, and with the other way not even close).
A pure between subject effect for example would be if the activity to all conditions in patients is different than in controls, so that within permutations have no effect anyway?
Best Wishes
Silvano
|