Dear Luis,
a few comments for the moment:
>> My hypothesis is that Pre A = Pre B, and Post A != Post B, which should translate in a significant interaction.
>> And that Post A > Pre A.
Not sure about the != , is this a ≠ (not equal sign)? If it reads Pre A = Pre B and Post A = Post B, this does not mean there's an interaction. You would have an interaction if (Post A - Pre A) - (Post B - Pre B) not equal to 0.
>> Are this statiscally valid contrasts, assuming that Post A and Pre A occur in different sessions?
Similar paradigms often incorporate a control condition which is assumed to either be stable across time or to undergo just some unspecific changes. You could thus add a third condition C and test for (Pre A - Pre C) - (Post A - Post C) not equal to 0. Maybe B already reflects such a control condition.
>> Pre A = Pre B
Pre A = Pre B implies testing of equivalence. Finding no sig. differences for Pre A - Pre B (or no sig. differences for Pre B - Post B) does NOT mean that they are equivalent, although such findings are frequently interpreted this way (e.g. "intact processing" in the context of patient groups).
Best,
Helmut
|