I’m doing gPPI by following the pipline here: https://afni.nimh.nih.gov/pub/dist/edu/data/CD.expanded/AFNI_data6/FT_analysis/PPI
In the cmd.ppi.2.make.regs part to make the regressors,there is one step generating the ideal IRF (starting from line 66) . There are two options in the demenstration, one using GAM and one BLOCK, with different durations.
My stimuli have varying durations, so when I do GLM, I used combination of -stim_times_AM1 and dmBLOCK. In the PPI it is said the basis function should be consistent across classes, so I suppose I should use dmBLOCK to generate an ideal IRF? But what duration should I use? Each block has different durations, from 15s -22s.
Sorry but the message is buried.
You should use the BLOCK basis function, as it matches what you are using with dmBLOCK.
The basis function used in the script should match the shape of your applied function, not the duration. It is supposed to be an “impulse” function, meaning it can be convolved to represent the response to longer events. Your dmBLOCK functions have the same shape, but then are additionally convolved based on the duration of each event. PPI will work in a similar manner.
The GAM and BLOCK durations, 12 and 15 seconds, represent the basic durations of the impulse functions, not the durations of events. If you plot a basic GAM function, it takes about 12 seconds to return to 0. The BLOCK function takes about 15 s.
Does that seem reasonable?