Hi Bonnie,
I would not be inclined to oversample the EPI data to match
the regressors; it seems hard to justify. Oversampling a
smooth function is one thing, but oversampling noisy data
adds a fairly big unknown.
It would lead to much higher single subject statistics of
course, but that would be due to artificially doubling the
degrees of freedom. Whether the betas would be more accurate
is hard to say.
It is fairly common to have regressors at a higher resolution
than the EPI data (and of course using -stim_times, we could
sample the data at pretty much any resolution). For example,
respiratory and cardiac data may come at 50 Hz. But we still
do not oversample the EPI data.
The EPI data only comes so fast, and that is the signal we are
trying to evaluate. You can certainly trick the programs that
way, but I would not recommend it.
- rick