Hi Pengmin,
It depends on what you want for the null hypothesis of the
contrast. As '1 -1 -1' the NH would be that the resopnse to
stimulus S1 equals the sum of stimuli S2 and S3. As '2 -1 -1'
it would be that S1 equals the average of S2 and S3.
The latter case is more typical, comparing S1 to the average
of S2 and S3. Note that in this case the contrast might be
written as '1 -0.5 -0.5', which is more precisely a comparison
with the average.
---
Doesn't 3dANOVA output FDR curves for -contr lines? I would
expect you to see the q-values in the afni GUI when that t-stat
sub-brick is chosen for the threshold.
- rick