Hi Christy,
Running 3dvolreg a second time should not gain you anything,
assuming you have already saved the "distances from the original"
via -1Dfile in your first execution.
Ziad's suggestion was not to look at the distances moved, but at
some semblance of the "error", computed as the difference of each
registered TR with the registration base:
3dcalc -a run1_registered+orig -b base_image -expr a-b -prefix diff_run1
This you could have afni run through in time to visually look for
any big differences. Or you could change the expression to
"-expr '(a-b)*(a-b)'" and run 3dmaskave on it, to look for large
squared differences quickly.
---
Between run motion should not be a problem, unless it is big enough
to affect the shimming. It is not necessary to censor time points
around a change in run. Those steps in the motion parameters will
be absorbed by the polynomial baseline in the linear regression.
Based on that, if you tell 1d_tool.py where the run breaks are, it
will _not_ count them in the derivative (i.e. the derivative will be
zero at the start of each run).
- rick