It seems to me another approach to normalizing individual differences for group comparisons in our unique case would be to SCALE, on a voxel-wise basis, these deviations from zero mean by dividing all values in the voxel by the absolute value of the MAXIMUM deviation that occured in the time series.
The following script seems like it would work, but does not result in a scaled 3d+time series dataset (on which to apply a regression).
3dTstat -abs(max) -prefix maxbrain par12dsbpfnew+orig
3drefit -abuc maxbrain+orig
3dcalc -datum float -a par12dsbpnew+orig -b maxbrain -expr '(a/b)' -prefix normalized
What is wrong with this picture?
Jim B