Hi Sam,

That is a good point.

When estimating the blur in the resting state data, it
probably makes sense to use the all_runs dataset (the
"uncleaned" time series - which has all of the noise).
Of course, in this context, one might argue about exactly
what noise means. But in theory, resting state data
should lack the strong BOLD signals that might have a
meaningful effect on the blur estimates.

Generally the all_runs and errts time series have very
similar smoothness levels, but I could imagine that
after the regression operation, especially with any
bandpass regression, the smoothness could be affected.

Because of that potential effect, I will adjust the
examples to use -regress_est_blur_epits.

If you go that route, consider using both options to
compare with. Hopefully that will be comforting
(assuming they are similar).

- rick