Dear Alfonso,
Dear All,

It may seem trivial... May I ask you what is the proper way of doing simple correlation analysis?

Setup covariates 2nd level:
I have two groups – drug and placebo, let's say each with 2 subjects. I code All 1 1 1 1, Drug 1 1 0 0, Placebo 0 0 1 1. I also have behavioral ratings, which need to be correlated with connectivity. Now, if you want to test just the drug group, do you put the ratings for all (drug value, drug value, placebo value, placebo value) or the group with 0 (drug value, drug value, 0, 0)?
Results contrasts: 0 1

Thank you.
Lucas

Dear Lucas,

If you create three covariates: 'Behavioral' (e.g. 1 3 2 5), 'Behavioral_drug' (e.g. 1 3 0 0) and 'Behavioral_placebo' (e.g. 0 0 2 5), then:

a) to correlate behavioral ratings with functional connectivity in the drug group only, select 'Drug' and 'Behavioral_drug' and enter a [0 1] contrast
b) to correlate behavioral ratings with functional connectivity in the placebo group only, select 'Placebo' and 'Behavioral_placebo' and enter a [0 1] contrast
c) to correlate behavioral ratings with functional connectivity in the entire group, select 'Drug', 'Placebo' and 'Behavioral' and enter a [0 0 1] contrast (we are using this, instead of the standard 'AllSubjects' 'Behavioral' with a [0 1] contrast, so that we can control by main group effects when computing the behavioral correlation of interest)

If, in addition, you want to look at the group-by-behavioral interaction (differences in the associations (a) and (b) above), you would use:
d) select 'Drug', 'Placebo', 'Behavioral_drug', 'Behavioral_placebo', and enter a [0 0 -1 1] contrast

Hope this helps
Alfonso
 
Originally posted by Lucas Moro: