> I assume this difference is due to differences in df when running the between
> group vs. within group analysis. Is this correct?
In addition to the difference in the degrees of freedom, GroupAna uses the pooled variance to calculate the statistics. In other words, the data from group B are also involved in computing the statistics for the contrast for Condition1 in Group A.
> Which way is the more appropriate way to describe the within group differences?
GroupAna uses more data in modeling, but also bears an assumption. For example, for this specific contrast, it assumes the variability of the contrast within group B would be roughly the same. If this assumption doesn't hold, the 3dANOVA3 result would be more appropriate.
Gang