You also need to demean your Data using "-D" option in Randomise.
> Date: Fri, 5 Oct 2012 04:39:08 +0100
> From:
[log in to unmask]> Subject: [FSL] tbss correlation
> To:
[log in to unmask]>
> Hello,
>
> I want to look at the correlation between 16 patients' FA vs their behavioral data. I'd like to know if my glm setup is correct. Under the Group column, I put all 1's, then I created 1 EV and filled them with demeaned behavioral scores. Two contrasts, 1 for positive correlation, -1 for negative correlation.
>
> My other question is whether to use the "-D" option when running randomise?
>
> Thanks,
> Amy