Hi Donald, and everybody else,
First of all, thanks for your answer.
I indeed used the interaction contrasts that you mentioned and those which are in unmistakably detailed in my message.
I still feel that something is missing:
If I'm not mistaken, the second contrast which indeed includes my answer represent four cases:
1) Voxels for which A1 > A2 for females and A1<A2 for males (both significantly)
2) Voxels for which A1 > A2 for females (significantly) and A1=A2 for males (i.e. the null hypothesis cannot be rejected regarding males)
3) Voxels for which A1 < A2 for males and A1<A2 for females (both significantly)
4) Voxels for which A1 < A2 for males (significantly) and A1=A2 for females (i.e. the null hypothesis cannot be rejected regarding males)
My question is which contrast (if any) could be used to isolate only the two first cases, which are the only true A1 > A2 for females. Of course I can create a separate model which includes only the females and then do a simple T test but I would be missing the Between Error of my whole study (including the males) for more accurate statistics.
|