Hi Everyone,
I've run a TBSS analysis, and have some very nice results, in FA, MD and RD. The data underwent 5000 permutations, TFCE and FWE correction. I was interested in taking a look at FDR correction maps, and when I ran it, the p-value at q=0.05 is 0.
I followed the steps on the FDR page...
just briefly, for instance:
fslmaths tbss_RD_tfce_corrp_tstat2.nii.gz -mul -1 1-tbss_RD_tfce_corrp_tstat2.nii.gz
fslmaths 1-tbss_RD_tfce_corrp_tstat2.nii.gz -add 1
fdr -i 1-tbss_RD_tfce_corrp_tstat2.nii.gz -m tbss_mask_3_skeleton_bin.nii.gz -q 0.05
And the result is 0.
I don't really understand how this is really possible, considering that in RD and MD, there are quite a few (very) large clusters, and the FA findings are by no means small...
Any ideas?
Tom Nichols suggested that perhaps the T-scores had been shifted from 0 (in a similar question elsewhere), and, if this is the case:
1) How does this happen
2) Can I fix it?
3) Does it 'take away'/delegitimize my FWE results?
Thanks,
Massih
|