Hi Carlos,

I assume that EV1 and EV2 are 1st level EVs (otherwise you wouldn't want the interaction). Then you can compute the interaction in a new model as:

newEV1 = oldEV1 - oldEV2
newEV2 = oldEV1 + oldEV2
newEV3 = RT
newEV4 = newEV1*newEV3
newEV5, etc = other EVs you may have

Then create a contrast that tests newEV4:
C1: [0 0 0 1 ...]
C2: [0 0 0 -1 ...]

Note, however, that the oldEV1 and oldEV2 need to have already been convolved with the HRF. Perhaps the best way is to simply copy them from the already generated design matrix from FEAT, then create a new model in which these already convolved EVs are entered (but don't convolve again then).

Hope this helps.

All the best,

Anderson


On 2 April 2016 at 23:10, Carlos Gonzalez-Garcia <[log in to unmask]> wrote:
Dear FSL users,

I've previously implemented a GLM analysis for our multi-session and multi-subjects design. Now, we are interested in one of the contrasts (e.g. EV1 > EV2), and I want to test if some measures (such as RTs) actually interact with this contrast.

Please note that I have a value (e.g. RT) for each trial, session and subject, so I think I have to set the interaction at the first level of analysis. I've tried to do so adding an additional EV to my first level model with the (demeaned) measure and setting a F contrast between this EV and the contrast of interest (EV1 > EV2). However, when I try to run a fixed effects analysis across sessions, I can't add the results of the F contrast.


My question therefore is if there is a way of testing this interaction in my desing.

Thanks in advance!
Carlos