Dear Noelia,
1) With [-1 2 -1] you test whether GM density/volume is larger at T2 relative to the average of T1 and T3. Assume some GM values like 0.5 1 0.8 or 0.5 1 1.1, then the test might turn significant, but it's probably not what you want to test.
2) If you want to find out whether the difference [-1 1 0] is identical to [0 1 -1], or in other words, whether the T1 is identical to T3, then you want to assess equivalence. Note that with the "default approach" you can't prove equivalence (proof of the null hypothesis), you may only conclude that there was no evidence for the alternative hypothesis. But this does not mean you "confirmed" that there are no differences (like you claimed for group 1 vs. group 2 at time point 1). This is a very common mistake though, in clinical context often interpreted as "intact function/structure in patient group x".
One form of equivalence testing is the "two one-sided tests" procedure. In its simplest form you test whether the absolute value of (T1 - T2) is at least as large as a critical value (null hypothesis) or smaller (alternative hypothesis). However, this is still very, very uncommon in psychology/neuroscience I'd say. For VBM studies at least it should actually be relatively easy to determine some critical value, based on other aging studies.
Best,
Helmut
|