Go ahead and add your other continuous covariates and demean them all as you add (even gender, so it will no longer be 0's and 1's). That way the [1 0 0 0...0] contrast will be the mean for group A for the average value of everything else. The [0 0 1 -1 0 0...0] contrast will not be impacted (much, if at all) by demeaning the other covariates. Of course don't demean the first two covariates (your GroupA/GroupB covariates) Cheers, Jeanette On Thu, Feb 7, 2013 at 11:26 AM, Jie Yu <[log in to unmask]> wrote: > Dear Jeanette > Back to my first question..... > Sorry, I just realise that I have other variables to control when trying > to find different correlation between group A and B. > age , gender and total gray matter.... > But I think I still don't need to demean any variables or use "D"? > Thanks > Jie-Yu >