Design matrix for longitudinal VBM

[araikes]

Apart from the issue I just posted (#184), I was hoping I could confirm how SwE is expecting the design matrix to be built. I'm doing a longitudinal VBM with two groups (treatment and control). I have an unbalanced design due to image quality and dropout. All of the preprocessing was done in CAT12 (current version as of the end of November).

Here's what I've set up in SwE (to parallel what it would be in SPM/CAT12's flexible factorial model):

• Modified SwE
• Groups: 54x1 double with (1=control, 2 = treatment) in order of scans
• Visits: 54x1 double with (1= baseline, 2 = post-treatment) in order of scans
• C2 with approx III df
• Subjects: 54x1 double with participant IDs. These aren't sequential (e.g., 2, 6, 15, 26, etc).
• Covariates: One covariate each for treatment_baseline, treatment_ptx, control_baseline, control_ptx (54x1 doubles with 0s or 1s).

Contrast (for example): F-test for group x time interaction [eye(2)-1/2 1/2-eye(2)]

Is that the correct set up when using SwE for this kind of analysis?

[nicholst]

• Modified SwE
• Groups: 54x1 double with (1=control, 2 = treatment) in order of scans
• Visits: 54x1 double with (1= baseline, 2 = post-treatment) in order of scans
• C2 with approx III df
• Subjects: 54x1 double with participant IDs. These aren't sequential (e.g., 2, 6, 15, 26, etc).

This is fine

• Covariates: One covariate each for treatment_baseline, treatment_ptx, control_baseline, control_ptx (54x1 doubles with 0s or 1s).

Looks good

Contrast (for example): F-test for group x time interaction [eye(2)-1/2 1/2-eye(2)]

This F contrast is redundant. Walking through the effects...

The main effect of time is

-1  1 -1  1

The main effect of group is

-1 -1  1  1

and then the interaction is just 1 row

 1 -1 -1  1

Make sense?

[araikes]

Great. Thanks. I just wanted to make sure that SwE was expecting it in the way I was conceptualizing it.