This isn't easily done. The gPPI code reads the covariates from the SPM.Sess structure. You'd need to modify that structure to contain the additional regressors and regressor names.

It is probably easier to re-estimate the model with the new covariates.

While I haven't done extensive tests, in my experience the white matter/CSF regressor haven't changed the results. Additionally, if you operate on the assumption that you need to remove variance from the white matter and CSF because it confounds the BOLD task signal, then one would argue that you should have removed it at the first-level to get a clean task signal.

Best,
Donald

Originally posted by Brianne Mohl: