Hi George,
Might it be worth considering adding '-stim_nptr 2' for this
in 3dDeconvolve? Then for run #2, instead of thinking the
stimulus is at time 6 seconds and the data is at time 7.5,
for example, the stimulus might be considered at time 7.5 and
the data at 9. That would leave the data TR locked, but the
stimuli would be 1/2 TR locked.
Another option would be to use TENT functions and -stim_times
(where the times would be sub-TR locked) in 3dDeconvolve.
This assumes that you want to do deconvolution instead of
regression.
Just pondering here...
- rick