Yes, as you listed, there are different ways to run such a group analysis.
One-way ANOVA (
3dANOVA) is not appropriate in this case because the correlation between the two conditions across subjects is not modeled.
For the contrast of A-B, paired
t test (
3dttest) or one-sample
t test on the contrast should be comparable to
3dAVOVA2 -type 3. However, the single effects of conditions A and B might be slightly different between the two methods because the assumptions under the two approaches are not the same. You can argue one way or the other, but I would simply run
3dAVOVA2 -type 3 since all the results come out with just this one shot.
See more discussion here: [
afni.nimh.nih.gov]
Cheers,
Gang