1. Measure
The Python engine computes PageRank, HITS hub and authority scores, betweenness, closeness and degree centrality with networkx, plus Louvain communities.
Loading…
Interactive knowledge graphs with statistical significance testing
The Python engine computes PageRank, HITS hub and authority scores, betweenness, closeness and degree centrality with networkx, plus Louvain communities.
The same graph is rewired many times with degree-preserving edge swaps. Every node keeps its exact in- and out-degree; only the wiring is random.
Each observed score becomes a z-score against its null distribution. A node is significant when it beats chance after a Benjamini–Hochberg correction at 5% FDR.
PageRank answers what is central, HITS separates collectors from sources, betweenness finds bridges between domains, and modularity tells you whether the clusters are real.
Degree centrality carries no z-score on purpose: the null model holds degree fixed, so degree can never be surprising. Significance here always means “more than this node's degree alone would predict”.