Dear all,
I repost this .. hope someone can help me ..
I'm using FDR correction in my statistics and I have some questions on
thresholding.
1. FDRill seems to work fine on the Niifti SPM5T maps, is it ok to use
it? (just changed the spm_get into spm_select) Someone else tried?
2. Can we use FDRill on SPM F maps? (the fact is for ex a cluster
survive some FDR corr with the F but does not for the corresponding t)
3. In addition to the paper on FDR, how can we explain that we use
'adaptive' thresholds with FDRill?
4. Although FDR remove false positives among activated voxels, does it
make sense to use an extended threshold? I still have small clusters of
1 to 5 voxels at 'unexpected' locations..
Thank you very much in advance
Best
Cyril
--
Pernet Cyril
Neuropsychologist, PhD
INSERM U455 Pavillon Riser
CHU Purpan 31059 Toulouse
France
Tel: +33(0)561779503
Fax: +33(0)561499524
http://membres.lycos.fr/wamcyril/
[log in to unmask]
|