Print

Print
Hi Jan,

For the question "Do the post-intervention means, adjusted for pre-intervention scores, differ between the 3 groups?", I understand that the pre-intervention scan isn't used, and that instead you have a pre-intervention clinical score, which you'd like to treat as a nuisance (i.e., adjust for it). Hence the "ANCOVA" in the subject line. Is this correct? If yes, then the pre-intervention scan would be ignored, and just the post-intervention scans would be used as the input file for randomise. The design would be:

EV1: controls (coded as 0 or 1)
EV2: intervention 1 (coded as 0 or 1)
EV3: intervention 2 (coded as 0 or 1)
EV4: pre-intervention score (mean centered).

The contrasts would be:
C1: [1 -1 0 0]'
C2: [1 0 -1 0]'
F1: C1 and C2

However, if what you really want is to see if the difference between post and pre intervention differ among the three groups, while controlling for a pre-intervention score, then the simplest is use fslmaths to calculate, for each subject, the difference between post and pre, and use this difference as the input for randomise. The design and contrasts would be the same as above.

Hope this helps!

All the best,

Anderson



On 22 September 2014 19:45, Jan B <[log in to unmask]> wrote:
Dear FSL experts,
after studying the FSL webpage it still remains unclear to me how to construct a design matrix in the following case (using GLM for further processing with randomise):

3 groups of subjects (control, intervention 1, intervention 2) underwent 2 fmri scans (pre-intervention, post-intevention).

The question is: Do the post-intervention means, adjusted for pre-intervention scores, differ between the 3 groups?

Does anybody know how to solve this problem with FSL?

Thank you very much for your suggestions!