When to choose dynamic vs. static social network analysis.

Abstract

There is increasing interest in using dynamic social networks in the study of animal sociality and its consequences. However, there is a general lack of guidance on the when and how such an approach will be valuable. The aim of this paper is to provide a guide on when to choose dynamic vs. static social network analysis, and how to choose the appropriate temporal scale for the dynamic network. I first discuss the motivations for using dynamic animal social networks. I then provide guidance on how to choose between dynamic networks and the "standard" approach of using static networks. I discuss this in the context of the temporal scale of changes observed, of their predictability and of the data availability. Dynamic networks are important in a number of scenarios. First, if the network data are being compared to independent processes, such as the spread of information or disease or environmental changes, then dynamic networks will provide more accurate estimates of spreading rates. Second, if the network has predictable patterns of change, for example diel cycles or seasonal changes, then dynamic networks should be used to capture the impact of these changes. Third, dynamic networks are important for studies of spread through networks when the relationship between edge weight and transmission probability is nonlinear. Finally, dynamic social networks are also useful in situations where interactions among individuals are dense, such as in studies of captive groups. The use of static vs. dynamic network requires careful consideration, both from a research question perspective and from a data perspective, and this paper provides a guide on how to evaluate the relative importance of these.