To: spm@jiscmail.ac.uk
Subject: S.V.C. with FDR
From: "Thomas E. Nichols" <nichols@umich.edu>
Date: Mon, 30 Sep 2002 14:48:54 -0400
FDR users,
Small volume corrected (SVC) inferences based on boxes or spheres
have not been working correctly with FDR. In particular, the
FDR-corrected p-values didn't change at all with SVC.
If you use FDR please download this updated file
http://www.sph.umich.edu/~nichols/FDR/spm_VOI.m
As an added bonus, your FWER-corrected p-values may be slightly
smaller. Previously, the SVC correction for a box or sphere was
based solely on the geometric volume, ignoring the analysis mask.
In this new version, voxels within the box or sphere but outside the
brain do not count as part of the search volume. This reduction in
volume affects the Bonferroni results but not the RFT results.
-Tom
PS: Thanks to Ahmed Toosy for reporting this problem.
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
nichols@umich.edu 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029