For these F-tests, all that matters is the range of the contrast matrix. That is, the actual vectors you stick in don't matter; what matters is the set of all possible linear combinations of them.
Put another way, if you have two such contrast matrices and the vectors(*) in one can be expressed as linear combinations of the vectors of the other, and vice versa, then they'll give exactly identical results.
(This is probably similar to Cyril's comment.)
(*): I say "vectors" because I believe we enter them as rows in the GUI, but in the SPM.mat file they're represented as columns.
|