Dai has a new paper in Animal Behaviour, co-authored with Damien Farine of the Max Planck Institute. The paper presents a method to assess the robustness of 'community assignment' procedures to detect social clusters in networks. The assignment of nodes to network communities is a popular method to understand the structure of societies and has implications for the evolution of social dynamics. However, there are some potential pitfalls when trying to use bootstrapping to assess the robustness of community assignments. We propose a way to combine bootstrapping with a measure of assortativity to determine whether community assignments are robust given a certain sample of observations. We find that the robustness of community assignment is contingent on both the sample size and the underlying structure of the network--e.g., more samples do not add robustness when networks are highly clustered or not clustered at all, but sample size does matter when community structure is intermediate (as is often the case).
Shizuka, D. and Farine, D. R. (2016) Measuring the robustness of network community structure using assortativity. Animal Behaviour. 112: 237-246. link