Hi all,
I recently received a manuscript review which stated:
"The primary cluster forming threshold of Z=2.3 is associated with an uncorrected p =0.010724 is considered liberal. A primary threshold of the image at p<.001 uncorrected has been recommended.
See Woo, Krishnan, Wager.(2014), Cluster-extent based thresholding in fMRI analyses: pitfalls and recommendations. Neuroimage. 1;91:412-9 "
Aside from the opinions of these authors, is there any thoughts from fellow FSL users and coders on loss of Gaussianity at Z=2.3 and if the default for Feat clustering should indeed change from 2.3 to 3.1 (i.e. 0.001 as in SPM), as recommended in this paper? I assume it depends on smoothness, and the default FWHM of 5mm is significantly lower than what is used by default in SPM, which sets up different smoothness estimation in the datasets… i don't know if perhaps this leads to different defaults in the two packages…
thanks for feedback on this!
-vitaly
|