Chris.

The interaction doesn't look like it was formed correctly. I would suggest using the flexible factorial model with 3 factors:

(1) Session (2 levels for the 2 measurements)

(2) Group (2 levels for the 2 values of the categorical variable)

(3) Subject (66 levels, one for each subject)

Then setup the models with:

Main effects for (1), (2), and (3) and the interaction of (1) and (2).

You can look at my prior posts on flexible factorial models and the contrasts to find instructions on how to generate any contrast. The search term: S1G1C1 should help find the previous posts.

Best,

Donald

Best Regards,

Donald McLaren, PhD

On Mon, Feb 8, 2016 at 7:01 AM, Weise, Christopher <[log in to unmask]> wrote:

Dera all,

I am trying to set up a paired t-test (VBM) with a total of 66 pairs. In addition I want to model an interaction with a categorical variable that does not differ within pairs (i.e. group status 0 and 1 for 43 vs. 23 pairs). I would like to determine whether within pair differences are affected by group status. However I'm slightly puzzled by the design matrix (attached) and how to set up the contrasts to answer my question. Any help is highly appreciated.

Thanks in advance & cheers

Chris