Dear Mailbase,
How delicate are the statistics involved in the random field corrections
(for z- and t-scores)?
I've noticed colleagues take one map, threshold it to generate a mask (not
of brain vs. non-brain, but of activation of some condition), then apply
this mask to another spm, then run a random field correction on the result.
Won't applying such a mask ensure that the resulting spm is *no longer* a
random field, invalidating the inference step?
Maybe it doesn't really matter; perhaps one could argue that some rule of
thumb applies in the choice of a corrected p-value, at least in the case of
the peak height test. But that's just a guess; and on the other hand, the
clever calculations near the end of Price/Friston ("Cognitive Conjunction:
A New Approach to Brain Activation Experiments", NeuroImage 5, 261-270
(1997)) used to justify steps involved in computing conjunction maps in the
context of random field theory indicate that one must be careful about such
things.
Best,
Stephen Fromm
NIDCD/NIH
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|