Hi Markus.
> sorry that i misunderstood your comments. do you know where i can find
> information about more sophisticated approaches and how to implement
> them in spm? especially the last design for non linear effects sounds
> interesting.
>
read back my e-mails and try the multiple regression approach - all con
images, 6 conditions, 100000, 010000, 001000, 000100, 000010, 000001
then simply perform your contrasts [1 0 3 0 5 0] linear gp1 [0 1 0 3 0
5] linear gp 2 [1-13-35-5] linear gp 1 vs 2 [-1 0 1 0 0 0] a<b gp1 [0 0
-1 0 1 0] b<c gp1 .... AND/OR as you already perform the a<b and b<c at
the first level try the multiple regression at the second level with 4
gps, 1000 (a<b gp1), 0100 (b<c gp1), 0010 (a<b gp2), 0001 (b<c gp2) then
test [-1010] U [0-101] for a<b<c (linear or not) different between gp1
and gp2
cyril
|