I've previously asked about the same finding. I never really got to the bottom of it because I
haven't had the time to investigate the suggestions that others gave me. Here's a set of
suggestions that may get to the bottom of this.
1) Make sure that you are using FLAME, rather than Simple OLS.
2) Ensure that you've designated the group membership correctly using the left-most buttons in
the feat GUI. I believe that these must be set for FLAME to model those variances correctly. See
"Group" column in http://fsl.fmrib.ox.ac.uk/fsl/feat5/detail.html#UnpairedTwoGroupDifference
3) Be sure that you have the most recent version of FSL (4.0.4). I was told that there was a bug in
some earlier versions that affected computing group variances.
4) I can't find the specific post, but previously on this mailing list, I saw someone orthogonalize
the group EVs (EV1 and EV2 in your case) within the feat GUI using the "Setup Orthogonalization"
buttons. Theoretically, it shouldn't change the results at all, but maybe they knew something that I
don't.
Good luck, and I'm sure that the list might appreciate an update.
--Greg
____________________________________________________________________
Greg Burgess, Ph.D.
Research Associate, Institute of Cognitive Science
University of Colorado - Boulder
Phone: 303-735-5802
Email: [log in to unmask]
University of Colorado - Boulder
UCB 594
Boulder, CO 80309-0594
|