Dear Stijn,
The two-way interaction corresponds to the difference of the differences, in your case (preA - preB) - (postA - postB). Thus you just have to set up a corresponding set of weights of 1 and -1.
Assuming a design matrix of e.g. pre_A pre_B pre_reg1 pre_reg2 ... post_A post_B post_reg1 post_reg2 ... pre_intercept post_intercept (with the "reg" columns corresponding to additional columns, e.g. other conditions, motion regressors, ...), then the F contrast vector would be [1 -1 0 0 ... -1 1 0 0 ...]. Replace the ... by additional zeros where necessary; a simple way is to go with zeros(1,n) with n corresponding to the series of subsequent columns that you want to weight by zero.
In your case, as it's two levels a factor, you can also just go with (directed) T contrasts, [1 -1 0 0 ... -1 1 0 0 ...] for (preA - preB) - (postA - postB) > 0 and [-1 1 0 0 ... 1 -1 0 0 ...] for (preA - preB) - (postA - postB) < 0.
Best
Helmut
|