Hi Joe,
This is not a bug.
Either way, align_epi_anat.py is used to align the anat to the EPI,
as you have seen.
But the example states that with the -volreg_align_e2a option, the EPI
will be transformed to align with the anat along with motion alignment.
It does not happen in align_epi_anat.py, but in the volreg block. The
reason for putting the transformations together is to avoid the extra
interpolation.
It may take reading the help example a couple of times for it to sink in
(and reviewing the resulting script).
---
In "Soup to Nuts" 3dMEMA.R is not used, 3dANOVA3 is. 3dMEMA.R uses
the t-stats in the group analysis, while 3dANOVA3 just uses betas.
So your guesses here look correct. I see no inconsistency with that
lecture.
The reason it is not so good to warp statistics is because they are
not (maybe assumed to be) normally distributed. While t-stats are
pretty close to this, other statistics are not.
The problem comes from interpolation. When a warped voxel gets its
value from the 27 nearest neighbors for example, those 27 values are
averaged in some way. But it is not terribly valid to average a bunch
of F-statistics.
---
At any rate, it's nice to see that your AFNI version is up to date
and that you are working to read through a lot of the help! :)
- rick
