Hello!
I am trying to get the peak magnitude of individual trials in a timeseries in order to correlate these trial magnitudes across trials between averaged timeseries of different ROIs. The timeseries are a result of a fast event-related design where the ITIs are randomly mixed between 8s and 16s with TR=2s. It seems it maybe possible to look at the timeseries itself and select out the 3rd, 4th and 5th timepoints of the trial (since these tps define the magnitude of the general IRF) following the onset of the stimulus (auditory) which is at most 180ms in duration. With 8s ITI the end point of the previous trial (where we assume a trial duration of 12s) coincides with the 3rd tp of the next trial. Obviously a trial duration is necessary for defining the deconvolution, but it is not clear that the HDR is zero at that point (actually in most cases it is not). However, it maybe many tps after 12s where there is essentially “little” response left from the previous trial. So my question is should we stick with looking at the TS and extracting out these points, or do we use what you guys have programmed with stim_times and “individual modulation” ? The TS way requires I think a way of removing the baseline but may not be as accurate as “IM” but apparently there was a message board message that said “IM” didn’t work with event-related design well. My question is how does “IM” work? I assume it is a deconvolution regression method and how appropriate is it in our case? What in general are your suggestions for our problem? I appreciate any help you can give.