Hi SPMers,
I'm doing a 2-sample T test comparing a single patient to a group of
controls. Not a brilliant experimental design I know, but I'm only a
medical student and I do what I'm told.
Anyway, when running the results, I've been selecting uncorrected 0.001.
This is because FWE gives me next to nothing, and FDR doesn't seem to work
right (FDRill shows that the T distribution is slightly positively
centred), however, uncorrected gives me significance in all the areas I'm
looking at.
Looking at the volume results table, significant voxels at uncorrected,
also show up as significant at p-FDR-cor (at around 0.001/2, etc). Whereas
p-FWE-cor shows highly insignificant at 0.9, etc). I read somewhere that
FWE can have problems when there are low numbers in 1 group.
Therefore, the question is, can I state that my results are significant as
p-FDR-cor is showing high significance and the overlays show as pretty
much what I had hoped?
Cheers
Richard
|