A generalised significance test for individual communities in networks

Area of Science:

• Network science
• Statistical analysis
• Graph theory

Background:

• Empirical networks often exhibit community structure, where nodes are densely interconnected within groups and sparsely between them.
• Communities within networks are typically heterogeneous in size, edge density, inter-community connectivity, and overall significance.
• Existing methods for analyzing community structure often lack flexibility in handling diverse community detection algorithms and quality functions.

Purpose of the Study:

• To develop a novel statistical method for assessing the significance of individual communities within a given network.
• To create a flexible algorithm capable of integrating with various community detection algorithms and their associated quality functions.
• To provide a robust framework for evaluating the statistical significance of network communities.

Main Methods:

• The proposed method quantifies the quality of individual communities using a defined quality function.
• Community detection is performed by optimizing the sum of these quality functions across all communities.
• The significance of each community is determined by comparing its quality function's distribution against that derived from randomized networks.

Main Results:

• The algorithm successfully estimates the distribution of the quality function for randomized networks.
• It calculates the likelihood of each community's significance within the given network.
• The method was illustrated and validated using both synthetic and empirical network datasets.

Conclusions:

• The developed statistical testing method offers a unique and flexible approach to evaluating individual community significance in networks.
• It accommodates various community detection algorithms and quality functions, enhancing its applicability.
• This method provides a reliable way to statistically assess the importance of communities in network analysis.