Dear André,
There are several constellations for sig. "differences between differences". If it's group 1 < group 2, this could be due to
- G1 showing no time effect ( t1-t2 = 0), but G2 showing a decrease over time (thus t1-t2 positive) -> 0 < pos.
- G1 showing an increase over time (t1-t2 negative), and G2 showing a decrease (thus t1-t2 positive) -> neg. < pos.
- both groups showing a decrease over time, but a larger decrease in G2 -> pos. < "more pos."
- ...
In your case, it's actually a group difference for a time difference (~ session difference) of a differential contrast (the difference within session), and when discussing the finding it might be relevant to refer to that difference within session. Thus you'll need some post-hoc test, or at least a plot of the estimates for G1 with regard to t1-t2, same for G2.
> However, although the pattern of activity was exactly the same, the corresponding statistics was different.
[1 -1] in model 1 has to be identical to [-1 1] in model 2 in case the con images of the 2nd model are the inverse of those from the 1st model. So it must be something like an incorrect vector for one subject or one subject accidentally excluded (or results obtained with an older and a newer SPM version).
Best
Helmut
|