Hi Gang and Rick,
Thanks for your very helpful responses. Just to confirm, I did the ANOVA both ways (with auc and with averaged magnitudes) and the resulting clusters are identical.
And I'll keep those subtleties about the scale factors in mind when it comes to interpreting the numbers that go into the analysis, although I see now why they don't make a substantive difference with respect to the results. ANOVA essentially involves looking at ratios of variances, and the scale factor doesn't change those ratios, so it doesn't change the results in any particularly meaningful way.
This thread has really helped increase my understanding of what is going on with HDR response modeling in 3dD and how it relates to group analysis using 3dANOVA*. Thanks so much for your collective patience.
-Joe