Dear Pei Ling,
In case your hypothesis is really A > (B + C), then the corresponding weights are indeed [1 -1 -1]. This is a valid T contrast. Something like [pi/4 -3 sin(7)] would also be valid with regard to statistics, but it might lack a reasonable interpretation in most instances.
Accordingly, [1 -1 -1] might or might not be useful when it comes to interpretation; likely it depends on the regions under investigation. E.g. when subjects are instructed visually then probably you observe similar activations in visual cortices for moving this or that or both sides. In contrast, it might well be interesting to test whether the combined task is larger/smaller than the sum of the single tasks in motor regions. In this context you might also want to look at Szameitat et al. (2011, http://dx.doi.org/10.1016/j.neuroimage.2010.07.069 ), even if their main focus is on dual tasks with auditory and visual stimuli.
Best
Helmut
|