Dear Andrew,

I run NBS to compare clinical and control groups, and I wanted to ask your advice with interpretation of the results. The results seem to be very sensitive to the selection of primary threshold. At threshold 3 it's just total mess (see the pictures below), and at the threshold 3.5 it's already no significant connections. All runs had contrast [1,-1], 5000 permutations, t-test, alpha 0.05.

Primary threshold 3


Threshold 3.3 


What could cause such effect? Is there something I could check to make sure that the results are not caused by some issue with data overall or with some specific participants?

Thank you so much for kind help and for the toolbox!

Best,
Liya

The threshold can be considered as the minimum meaningiful effects size that is of interest. It is certainly possible that increasing the threshold results in fewer connections declared as signifciant, since the weaker effects are thresholded away.

It may be that the magnitude of differences varies considerably across connections and thus as the threshold is increased, weaker effects are no longer captured