Dear SPMers,
I would really appriciate if someone could help me with the following VBM correlation contrast problem:
I have VBM GM results in two groups. Performing individual correlations with a bahvioural measure in each group, I got convincing results with widespread GM pos. correlations.
Now I would like to know where the correlation is stronger in group 1 than group 2.
Refering to https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=SPM;50b22fa0.06
(Re: VBM: multi-group correlation and interactions) I created a full-factorial model with
1 factor G (2 levels, group 1 and 2 incl. the VBM structurals) and 1 ALL-covariate (C) with all the behav. scores and interaction with Factor 1 and centering Factor 1 mean. This results in a four columns design matrix (G1 G2 C1 C2).
However, contrasts 0 0 1 -1 (correlation G1 > G2)
0 0 -1 1 (correlation G1 < G2)
do not show any significant results at all.
What did I make wrong? I am really confused and thank for any help.
Best, Robert
|