Hi Gang,
Let me make sure I'm understanding you (and you're understanding me)...
My input to waver is a .1D file with a column of 0's and 1's, where I have put a "1" for each time point during which the stimulus was applied. Because each stimulus presentation occurs for a period that encompasses more than 1 TR, the gamma functions will overlap and thus summate (as Rick has pointed out).
I then use this waver output, that represents the response during the entire stimulus presentation, as my stim_file to 3dDeconvolve. You are saying that the coefficient that comes out of this is meaningful (since my time series is in % signal change already) and does *not* need to be scaled even though the peak of the function that I'm using as my model (the waver output) is > 1?
Maybe I can understand it this way - would (almost) the same thing be accomplished by inputting a single gamma wave, that was modeled for 1 TR (and, thus, had a true peak of "1"), and then use a min_lag and max_lag in 3dDeconvolve so that the single gamma wave would end up encompass my full stimulus presentation time? How would those coefficients (added up) compare with the single coefficient obtained by using the full gamma wave model, that covered the entire time period of the stimulus, as the stim_file input?
Hope I haven't further confused the matter.
Thanks so much - Liz