Dear Jack,
Can I just follow up Christian's response and note that your experienced
with cluster statistics is fairly typical. The thing to remember is that
the choice of the height threshold (z-thresh) is entirely arbitrary but it
dictates how the p-value works. So for a low Z (2.3) and a low p (0.01),
you are looking for spatially contiguous clusters with very large extents
(often many hundred voxels). Obviously this can be anatomically
meaningless in many cases, and it sounds as if that's what you're
seeing. Try raising the Z-value to a more reasonable level (3.1, 3.5, 4.0)
and changing the p to 0.05 (the standard alpha level). If the "activation"
is there across auditory cortex, you'll see it using appropriate thresholding.
Cheers,
Joe
--------------------
Joseph T. Devlin, Ph. D.
FMRIB Centre, Dept. of Clinical Neurology
University of Oxford
John Radcliffe Hospital
Headley Way, Headington
Oxford OX3 9DU
Phone: 01865 222 738
Email: [log in to unmask]
|