Hi Gang,
For SPMG1, if the parametric modulation is related only to the regressor S, and I still want to see the contrast S-I, while for S is considered the parametric modulation effect, S[1]-I[0] shouldn't be an option? Taking a closer look at the experimental design, I am not convinced that the modulation effect can be attributed to the regressor I, so probably it will not make sense to estimate the slope effect for I. Similar question is for the contrast S-B: S[1]-B[0] could be considered if only for the regressor S is considered the parametric modulation effect, and not for the regressor B?
Function of your answer to SPMG1 case, I will see what are the appropriate contrasts for the SPMG2 case.
Thanks,
Ruben.