Dear all,
this must be a basic question for must of you, but
I need your qualified opinion.
I am running an spm analysis with a cannonical 2 by
3 design, and I wonder which of these two approaches is the
appropriate:
After specifying the matrix design with six
conditions (i.e., A1, A2, A3, B1, B2, B3) I might run for the first level
analysis:
A1: 1 0 0 0 0 0
A2 0 1 0 0 0 0
. 0 0 1 0 0 0
. 0 0 0 1 0 0
. 0 0 0 0 1 0
A6 0 0 0 0 0 1
which represents the averages on each condition per
subject
at the second level, then, I might contrast, for
example, A1>B1 with the *com images in a full factorial anova
the second approach would be doing direct contrasts
at the first level:
E.g, A1>B1 1 0 0 -1 0 0 per subject, and
then, performing per-specific-contrast one sample t tests at the second level
with all the respective subject's com images.
which of these two approaches is the most
appropriate? I think the second one is more sensitive per specific comparisons
while keeping the advantages of the random effects test. is not it?
thanks a lot
Roberto
