In the previous post, I simulated clustered networks with varying strength of within-group, between-close-groups, and between-group social relationships. In part 2, I am using some of the methods provided by the igraph package to test how well these simulated clusters can be (re)-detected.
2. Determine clustering using different methods and compare the results 2.1 Prepare R rm(list = ls()) library(tidyverse) library(tidygraph) library(ggraph) library(igraph) In addition to these packages, this notebook requires the installation of the following packages: DT, cowplot, ggrepel.
Social networks often exhibit some kind of clustering (or community structure), such as distinct social groups in animal societies, or kin groups (or families) within social groups. Individuals within such clusters are more likely to interact with each other than individuals from different clusters.
There are many algorithms to detect clusters in social networks, and one might work better than another under some circumstances (see, e.g., Emmons et al. 2016).