Regarding my AFNI Jedi skills.....though I won't be sitting on the jedi council anytime soon, I do know just enough to get myself into trouble.
If I understand correctly, then Bob's suggestion to model the individual coefficients with polort -1 and Gang's suggestion to use the divine GLT should yield the same results no? That is, in both cases each level is being modeled independently and the coefficients are subsequently tested for a quadratic relationship.
If so, it seems easy enough and I won't need to test my powers of persuasion by trying to convince Bob to add '_AM3' functionality.
andrew