I have a naive question about the validity of comparing PPI connectivity measures over multiple sessions.
I have 2 conditions (A and B) in 2 sessions (within subject design) and I would like to test whether connectivity from a specific seed region is different
1) between conditions(A_sess1 and A_sess2 > B_sess1 and B_sess2),
2) between sessions (A_sess1 and B_sess1 > A_sess2 and B_sess2)
3) interaction between session (sess1 and sess2) x condition (A and B).
To do this, instead of creating a PPI.ppi using condA [1] vs condB [-1] for each session as the interaction terms, I have made a main effect of condA in sess1, condB in sess1, condA in sess2 and condB in sess2 separately by putting 1 to one condition and 0 to the other to create the PPI regressors and used PPI.P, PPI.Y and PPI.ppi in the single subject GLM model. Then on the second level, I fed the con_images of each of the PPI.ppi contrast into the 2nd level analysis in a 2x2 factorial design [A_sess1, B_sess1, A_sess2, B_sess2] and I have tested for the main effect of condition [A>B:1 -1 1 -1], session [sess1 >sess2: 1 1 -1 -1] and the interaction [1 -1 -1 1].
In this way, I am aware that the PPI.ppi regressor output on the single subject level is not actually showing the interaction effect. My question is, whether it is still valid to test for main effect and interaction effect in this way.
Regards,
Atsuko
|