Hi Gang,
Sorry to butt in here, but I think you were right in the
first place.
If Deconvolution is done via the GAM curve (as is
typical), the shape of that result would be a neuronal
curve that already includes the stimulus duration. For
example, one might expect the deconvolution of a 21s
BOLD response to be a 9s boxcar for a 9s neuronal
response, which ought to match a 9s stimulus event,
say.
The re-convolution step would still use GAM and would
not be altered to accommodate the 9s stimulus event.
- rick