Dear NBS experts, 

Hello all, I am performing NBS for group comparison between high group and low group with 3 covariates (age, sex, neuropsychological test). 
The results turn out great, but it seems quite sparse in a way, because it shows multiple significant networks, total of 10! 
With t-threshold of 4.0, it gives 4 networks (I attached the screenshot of the result page (UPDRS_FA_40_LtDom_network#.jpg))
With t-threshold of 4.5, it gives 10 networks (attached the results(UPDRS_FA_45_LtDom_network#.jpg)). 
Few networks show consistent clusters, but some show a one edge as a network. 
I know it is possible outcome, but still wonder how to manage and to interpret this result.
How can this be possible? In what algorithmic logic? 

Thank you in advance. 

I will wait for your the reply!

Hi YaeJi, 

the NBS is most sensitive to detecting effects that form interconnected networks. However, it can detect individual edges, as you have found.

I suggest investigating the distribution of connectivity values for each of the edges comprising the networks found to ensure that outliers are not driving the effects. 

However, it is certainly possible for the NBS to detect single edges and this does not indicate that something has gone wrong. 

The algorithmic logic is that a single edge is the most simple form of a subnetwork. 

Best wishes,
Andrew


Originally posted by YaeJi Kim: