Maximal Clique and Graph Coloring Heuristics for Overlapping Community Detection in Network Psychometrics

Abstract

The study of graph coloring is especially timely given that some of the most popular models and methods for estimating symptom networks have theoretical ties to graph coloring. We use graph-coloring algorithms to generate all possible colorings of the complement of a threshold graph associated with a symptom network. This facilitates the identification of symptoms that have flexibility in their color assignments, which often indicates that they serve as a bridge between multiple cohesive subsets of symptoms. Two published symptom networks are used for demonstration purposes.