afni.nimh.nih.gov

Hi Steve,

Yes to both.

The detrending to produce the residuals is weaker than
the detrending to estimate the blur. As you might have
in your proc script, the blur estimation for the residuals
(as well as the regression input) is done using -detrend.

Also, blur estimation is done per run, again because of
detrending. Longer time series detrend at a higher
level (as would apply to the regression step). Detrending
per run is at a much lower level, and is more consistent
with the regression.

- rick