The ComDet Package - Iterated Community Detection for the Analysis of Community Membership, its Stability and the Detection of Fuzzy Community Boundaries.

Abstract

In recent years, network analysis has entered psychological science. In psychological network models, nodes are variables and edges are estimated relationships between these variables. In this project, we answer recent calls for the use of established network community detection algorithms to detect clusters in psychological networks. This technical note introduces the ComDet package (under development for use in R), a package aimed at extending the functionality of the commonly used non-deterministic community detection algorithm spinglass. ComDet repeatedly performs the spinglass community detection algorithm (based on the igraph package in R) for a specified number of runs, quantifies and summarizes the results across runs, and visualizes those results. Our extended community detection function builds upon the prior literature in two ways. First, it offers summarizing output that communicates the stability and variability of particular community results. Second, because this package is implemented in R, the same platform commonly used to assess psychological networks, it can be readily applied as part of the growing toolbox for assessing psychological networks. This technical note will be updated with new insights in the near future. The final aim of this project is to provide researchers with a freely available R-package that runs the ComDet function using spinglass and, potentially, other non-deterministic community detection algorithms.

A topography of 21 phobic fears: network analysis in an epidemiological sample of adult twins

“…Fourth, to establish the number and nature of groups of phobic fears, we utilized IComDetSpin (Werner, 2018), a recently developed procedure that uses the well-established spinglass algorithm and summarizes the results across (in our case, 5000) iterations (Werner, 2018). Spinglass is a non-deterministic community detection algorithm that has been shown to perform well in retrieving the true number and structure of communities (Yang, Algesheimer, & Tessone, 2016).…”

An Exploratory Network Analysis of Sexual and Relationship Satisfaction Comparing Partnered Cisgendered Men and Women

“…A community detection analysis was made as an additional way of exploring the structure of the networks. Connectivity between nodes was explored using the ComDet-function 23 , which is an extension to the igraph-package 24 that allows for running many iterations.…”

Primjena mrežne analize u psihologijskim istraživanjima

“…algoritam spinglass koji pronalazi klastere tako da broj i ponderirana snaga bridova unutar klastera budu veći od broja i snage bridova među čvorovima različitih klastera (Heeren i McNally, 2016). Budući da je taj algoritam po prirodi nedeterministički, odnosno rezultati se mogu razlikovati u više iteracija (Werner, 2018), u ovome je radu izabrano najstabilnije rješenje dobiveno u 1000 različitih iteracija.…”