Hi,
I have two groups (treatment, placebo) and I want to compare these group's resting state (per session). 
I found two options and was wondering whats the best way:

In setup 1st level analysis:
create a contrast with treatment as 1 and ctrl as 0 (per subject). 
create a contrast with treatment as 1 and ctrl as -1 (per subject).


In second level results:
select allSubjects and only treatment (contrasted 1 and 0) and then choose simple effects.
OR
select the subject to subject contrast that was done in the 1st level analysis (1 and -1)


Whats the best practice?

Thank you for the help and the toolbox itself.

Or D.

Dear Or,

Assuming this is a simple unpaired test, the correct way to approach this is to define 2 second-level covariates at the setup step: Treatment and Placebo.

In each one, you would assign a 1 for the treatment type corresponding to each subject. For example, assuming the following:

Subejet1 - Treatment
Subject2 - Treatment
Subject3 - Placebo
Subject5 - Treatment
Subject6 - Placebo

The 'treatment' covariate would be defined as [1 1 0 1 0] and the `placebo` covariate as [0 0 1 0 1].

Then, in the 2nd-level analysis (seed-based, voxel-level etc.) you should select both groups in the between-subject column. Then, assign them with the desired contrast (Treatment>Placebo or vice-versa) by typing in [1 -1] or [-1 1].

Hope this helps. Good luck!

Sincerely,

-Yuval.

You can also look it up in the CONN manual,

Originally posted by Or Duek: