Hi all
I was browsing through the message board and following some thread I came across a reply by Doug Ward that argues for using the max value of the convolved input stimulus function as the correct scaling factor:
[
afni.nimh.nih.gov]
Now, I was pretty sure that if we want to express the magnitude of a regressor coefficient in terms of percent change wrt the baseline level of the BOLD signal, the only important parameter was the original amplitude of the convolving function (GAM, WAV or else), NOT the max value of the convolved timeseries (which depend of course on the event temporal distribution). A previous lengthy thread (Liz, Gang, Rick) seemed to confirm this, but Doug's e-mail challenges my certainty (surely, by virtue of some authority principle...
any comments?
giuseppe