Dear all,

Recently I used NBS v1.2 to detect the significant components and found that some generated components comprising only one edge. This seems strange according to NBS algorithm. Because according to my understanding, the components with only one edges should not pass thresholding. 
Thus I need your help to clarify that it is the correct result? And if it is not correct, should I abandon those components comprising only one edge? 

Thanks in advance,

Qing Ma

Hi Qing,

it is possible to find significant components that comprise only one edge. This suggests that for most samples from the null distribution, no edges passed the initial test statistic threshold, and thus the maximum component size was 0 for these cases.  So a single edge that the passes the threshold in the real data becomes significant.


I would not abandon significant components comprising only one edge, but they merit further investigation. For example, the effect at these edges might be driven by a few outliers.

Andrew

Originally posted by qing ma: