Vinny,
The contrast you obtained with 3dANOVA2 is essentially a one-sample t test on Group B (which means you can run the same test with 3dttest), and that is why the DF is 12. However, the same contrast in GroupAna is done based on both groups with pooled variance. In other words, the variability is estimated with all 31 subjects, even for such a contrast focusing only on one group, leading to the DF being 29. That was also adopted in 3dANOVA2/3 before the modifications a couple of years ago. The new approach in 3dANOVA2/3 is slightly more preferable in case the groups have difference variances, but such a modification is a little difficult to implement in GroupAna. Plus we have 3dLME which can handle such a situation now.
Gang
