Hi Jackie,
Daniel and I were pondering a few points to make:
1. Note that unless you really want EPI data aligned to the
anatomy, it is safer to align the anatomy to the EPI data,
via "3dWarp -oblique_parent".
One reason for this is simply not to interpolate the EPI
as many times. Another reason is you will lose the slice
timing data when rotating the EPI data. Running 3dTshift
first (if you are going to run it at all) will alleviate
this problem.
2. Are you still using 3dAllineate? I would expect not, since
applying the oblique transformation to the anatomy should
give you a good alignment.
If the alignment is not good, you may want to stick with
3dAllineate (to align the anat to the EPI), instead of using
3dWarp.
3. You should not be trying to align the statistical data with
the anatomical. The EPI and anatomical data should be aligned
to begin with. Any resulting statistical data will therefore
be aligned, too.
4. The quality of alignment should be viewed by overlaying the
EPI data on the anatomical data, not the statistical data on
the anat. The statistical data might be created from smoothed
EPI data, for example, which might make it show up outside the
brain.
- daniel and rick