Hi Dylan, 

these two approaches won't necessarily give you the exact same result, but they are both reasonable approaches. I would probably start with the computing differences a priori - second approach.  

Neither approach will tell you whether change in connectivity "predicts" the behavioural score. They will reveal whether there is a statistical association between change in connectivity and behavioural scores. 

Andrew

Originally posted by dylan sutterlin:

Hi,

I have a repeated measure experiment (pre, post intervention), and I would like to test for changes induced by the intervention, and if those changes are associated with a between-person behavioral score.

I previously found two networks that have differential weight after intervention with the following example design matrix and contrast (ex. for 3 subjects) : 

1 0 0 1
0 1 0 1
0 0 1 1
1 0 0 -1
0 1 0 -1
0 0 1 -1

Contrast = [ 0, 0, 0, 1] 

1) To test if a between-person score predict these changes, is the interaction the only way of doing that ? That would mean that my main test would now be on the interaction column of the intervention * behavioral score. Then I would have no "main effect", rather only an interaction for my question? 

2) The alternative to that would be to manually compute the connectivity matrix of the change between z_scored(post_intervention_adj)  - z_scored(pre_intervention_adj), then to test the main effect of the behavioral variable on the connectivity.

Should those two option be equivalent? I am very greatful for any advice.

Thank you for this tool!

Best,

Dylan