Dear Andrew,

I’m running the NBS for a two-sample t-test and have used two separate contrasts (positive and negative effects). My edge values are Fisher-z–transformed correlations. I’m consistently finding increased connectivity in the disease group, which is unexpected, and I wanted to confirm:

1. 
   

   Does NBS ever take the absolute value of edge weights internally, or is the sign preserved throughout thresholding and component formation?

   
   
   
2. 
   

   Are there key checkpoints you recommend to ensure the results are correct (e.g., verifying nbs.GLM.y against original data, checking matrix symmetry, and appropriate primary threshold selection)?

   
   
   

Thank you for your time.

Best,

Asli