You can get averages by taking any ROI dataset
(that you make or that simply exists, such as the
CA_EZ_MPM atlas, just to throw one out there)
and running 3dROIstats with that mask and your
result dataset. That will output an average for
each ROI.
Regarding what are reasonable p-values, it depends
on your actual steps. If it is from a seed time series
multiplied by 1/0, then very high values are expected
(e.g. at the seed). If you are multiplying by 1/-1/0,
then it is harder to say.
- rick
