Dear SPMers,
I would appreciate any pointers or comments on the following question: How should binary variables be coded in the design matrix in an fMRI paradigm (e.g. 0 and 1, -1 and 1, etc), IF, in the contrast one is running, gender is simply a nuisance variable and set to 0?
My question is based on the logic behind centering continuous variables, e.g. centering age so that the contrast [1 0], where the first variable is activation and the second variable is age, yields the average activation for individuals of mean sample age (because the mean for age is now 0).
But how does that work for binary variables such as gender? If I coded males as 0 and females as 1, and then ran the same contrast as above [1 0], would that then be the mean activation for males only? In other words, would it make more sense to code males as -1 and females as 1 in order for the contrast [1 0] to make sense and yield mean activity for 'mean' gender? Of course, I realise that there is no such thing as mean gender, but for the interpretation this might mean something like the average individual, i.e. regardless of gender?
If so, I guess the same would apply to running a multiple regression analysis in the context of VBM analyses?
Thanks in advance, any advice on this would be super helpful and much appreciated!
Regards, Loreen
|