As I understand it, SPM2 gives you four columns:
* group 1
* group 2
* ANCOVA regressor for gender
* constant
(Aside: you might want to consider doing a full group X gender analysis. That is, you're trying to model the main effect of gender, but in doing so you're ignoring the group X gender interaction. I believe "conventional" textbook stats would say that you can ignore the interaction only if it's small.)
Mathematically, it doesn't matter whether the constant column is there or not. The two models are equivalent. One can argue that having the constant column makes coming up with the desired contrasts confusing, but it's not too bad really.
You could tinker with the SPM2 code to eliminate the constant column, but it'd be easier just to get the contrasts correct with the constant column there.
|