Dear Paul,
> After reading the manual of VBM8 toolbox, i found i can set the statistical criteria like "uncorrected p value < 0.001 with expected number of voxels per cluster (which derived from SPM result table)". Is this kind of statistical criteria considering multiple comparisons correction in cluster level statistics?
I believe that the "expected number of voxels per cluster" is based on
the smoothness of your data and random field theory—that is, if
everything in your data is random and gaussian, you will still get
some overlapping clusters, and that is the expected size of them. For
a significant result, you would want more than this. I am not 100%
sure on the calculation of the expected size, though, so hopefully
someone will correct me if I'm wrong.
For cluster level correction, you'll want to look in the column of the
results table that gives you a corrected p value for the cluster.
These can be based on either random field theory (labelled FWE) or
FDR.
Of course, other methods (e.g. permutation methods) can also provide
valid cluster level p values, they just don't show up by default in
the SPM results table.
Hope this helps!
Best regards,
Jonathan
--
Jonathan Peelle, PhD
Department of Otolaryngology
Washington University in St. Louis
Office: (314) 362-9044
http://peellelab.org || http://jonathanpeelle.net
|