Hi all,
Stefan Kiebel wrote:
> Because the images enode probabilities you should use an absolute
> threshold to discard low probabilities.
Presumably the relative threshold will also discard low probabilities?
What are the pros and cons of absolute vs relative thresholding?
Does anyone have any comments/suggestions on which to use, and at what
levels -- abs 0.05 and rel 0.8 seem common, but what was the basis for
this choice originally? Is rel 0.8 generally higher or lower than abs
0.05 for a typical set of smwc1 images?
The SPM5 help for relative thresholding states that it's a "proportion
of the global value" -- is this the global mean across all voxels of
all images, or something else?
How does Tom Nichols's comment (http://www.sph.umich.edu/~nichols/NIpub/)
"Thresholded statistic maps can be seriously misleading"
relate to all this??
Many thanks,
Ged.
|