Hi Miguel,
I'll let someone else comment on the FDR vs. FWE issue, but a q value of
0.6 is way high. At q=0.6, 60% of the voxels with p-values less than
the determined threshold would be expected to be false-positives voxels.
q-values of 0.05, or perhaps 0.1, are much more reasonable.
cheers,
-MH
On Wed, 2010-12-22 at 15:40 +0100, Miguel Burgaleta wrote:
>
> Hi FSLers (and more specifically T. Nichols, perhaps?),
>
>
> I am running into a strange result when using FDR to correct my
> *vox_p* output from TBSS. I get surviving voxels with FWE (P<0.05),
> but not with FDR. This happens with all of my p-maps. Does this make
> sense? I have found a couple of unanswered posts pointing in the same
> direction.
>
>
> I am feeding FDR with the same mask that I used in randomise (just in
> case it matters). Played around with the -q threshold and started
> getting non-zero answer when setting -q 0.6 or above...
>
>
> Thanks in advance for any comments,
> Miguel
>
>
>
>
>
>
>
|