Hi, and thanks for this nice toolbox.

I'm working with tractography data from FSL's probtrackx. I usually apply a low threshold to get rid of spurious connections, based on the group-average connectivity matrix.

In the paper describing NBS, you propose to work directly with raw data in order to avoid systematic sparsity differences. In my case, some NBS networks could get significant even though they mainly consist of connections that I would normally not consider, as they are quite unlikely to be true.

What strategy would you advice ?

Thanks,
David