Dear all!

We are currently working on a VLSM-analysis in n=100 subjects but with rather small lesion volumes. Thus, there are only few voxels with a comparatively high percentage of subjects with a lesion in the respective voxel.

We used normalized data (t-scores) as behavioural data input und computed the VSLM for binary maps and continuous data using Brunner-Munzel test.

The *.txt-outputs (for different sets of behavioural data) show 

- rarely significant results for "+FDR" (FDR-corrected); however, here the result is >0 and the the thresholded statistical map seems correct
- far more often no result for "+FDR" but a negative value for "-FDR" (FWE-corrected).

First of all, I am surprised to find significant results with FWE but not with FDR-correction. I assume this might be due to our data distribution with only a small percentage of affected (lesioned) subjects by voxel (?)...

Second, we are uncertain about how to interpret the negative value (i.e. output of "-FDR"). It does not seem to be appropriate (and gives odd results) when using the negative value as lower bound when thresholding the statistical map.

Thanks a lot for your help!

Best,

Carol