> I agree that what you want to measure is the smoothness
> of is a residual map from 3dANOVA.
> Obviously, the residuals are in the program.
> Someone has to set them free.
If the error term for each contrast is what you want, it is free already in the output of 3dANOVA2 or 3dANOVA3. It is embedded in the corresponding statistical subbrik (usually
t), and a little bit manipulation with the intensity subbrik with
3dcalc would generate the error file. However, do realize in a within-subject (repeated-measure) ANOVA, such an error term might be different corresponding to each contrast, and there is no such a thing as universal residuals for all contrasts in this kind of design.
Gang