> I’m trying to create an interaction map (A1 > A2) > (B1 >
> B2) that will show the direction of (positive and negative)
> effects.
It seems to me that a more straightforward way is to run 3-way ANOVA instead of 2-way.
> Can this be done using 3dANOVA2? For example,
> using 3dANOVA2, I can produce separate maps with
> direction by labeling the 2 X 2 design as 4 conditions
> (A-D), and creating the contrast for A > B using –acontr 1
> -1 0 0, and for C > D using –acontr 0 0 1 -1. Given this,
> can I also specify (A > B) > (C > D)? If so, how?
If the 4 levels are A1 A2 B1 B2 in 2-way ANOVA, then the hypothesis "A1 = A2" (not A1 > A2) is 1 -1 0 0, and 0 0 1 -1 for "B1 = B2". For (A1 - A2) = (B1 - B2), it is 1 -1 -1 1. Then you can figure out the direction of each contrast based on the sign of either its intensity or
t value.
Gang