Our group conducted a event-related fMRI experiment. A little question in first-level GLM construction confuses us these days:
Each trial in our experiment contained a cue screen (2 sec.), several decision points (5 sec.) and several inter-stimuli-interval (ISI, 2 to 5 sec. randomly) between decision points. The ISI was a white cross presented on a black background. Should we include the timeseries of the ISI in the first-level GLM?
We are pleasure to hear your advice. Thanks in advance!
I would not model the ISI, as it should correspond to the implicit baseline of the experiment. Including it would might degrade the conditioning of the design matrix.
Thanks for your reply. The idea in FSL course - first-level analysis also recommends not to model the null events in the GLM.
Could I ask a follow-up question if you don’t mind? We are interested in the effect of decision points in our experiment, not the cue screen. Should I model the timeseries of cue screen in the first-level GLM?
Yes, i would model the cue, to avoid the cue effect being confounded with the decision condition.
Got it, thanks for all your help : )