dear list,
this morning my SPM2 starting doing some things I have never seen
(or noticed) before when I selected the FDR correction for a one-sample t-test rfx analysis on
con* images. for one comparison, the FDR height threshold (as reported
in the output window) was t=2.41. That's OK, but the uncorrected p<0.001
was t=4.79. shouldn't the FDR be stricter than the uncorrected? then, to
make things stranger, for another comparison the t-threshold for FDR was
Inf, but for FWE it was 17.1.
my (normalised) voxel size is 2x2x2, there are 64x64x24 voxels, and my
smoothing of the original images was with a 4.5 x 4.5 x 13.5 FHWM.
presumably I've done something strange somewhere along the road to this
analysis: any ideas what it might be?
thanks very much,
chris
Christopher Summerfield
[log in to unmask]
|