I am using FSL to model my fMRI data of a GLM with multiple parametric regressors. I post my design matric in below, please help me to check it out if it is decent to test the modulations in what it supposed to be.
There are 4 regressors in my GLM (say, A, B, C, D), and regressor A was modulated by 4 parameters (say, P1, P2, P3, P4). The GLM I built was Y = A + A(P1) + A(P2) + A(P3) + A(P4) + B + C + D. In Feat’s full-model setup menus, I used 5 EVs to represent regressor A and its parameters.
Given that FSL only allows 1 parameter in 3-column file, EV1 was indicated by the time series of regressor A and the 3rd column was all 1 which played as the constant. EV2 to EV5 were indicated also by the time series of regressor A, but the 3rd columns were modified to the specific values of P1 to P4 (refer to this). In additional, I also orthogonalized EV2 w.r.t EV1, EV3 w.r.t EV2 and EV1, EV4 w.r.t EV1 to EV3, EV5 w.r.t EV1 to EV4 (refer to FSLPPI wiki page - orthogonalization).
In contrast setup, I set the main effect of every parametric regressor to 1, like
EV1 EV2 EV3 EV4 EV5 …
0 1 0 0 0 … P1 effect
0 0 1 0 0 … P2 effect
0 0 0 1 0 … P3 effect
0 0 0 0 1 … P4 effect
Any idea is welcome, many thanks to your time.