Inferring animal social networks with imperfect detection

Abstract

Social network analysis provides a powerful tool for understanding social organisation of animals. However, in free-ranging populations, it is almost impossible to monitor exhaustively the individuals of a population and to track their associations. Ignoring the issue of imperfect and possibly heterogeneous individual detection can lead to substantial bias in standard network measures. Here, we develop capture-recapture models to analyse network data while accounting for imperfect and heterogeneous detection. We carry out a simulation study to validate our approach. In addition, we show how the visualisation of networks and the calculation of standard metrics can account for detection probabilities. The method is illustrated with data from a population of Commerson's dolphin (Cephalorhynchus commersonii) in Patagonia Argentina. Our approach provides a step towards a general statistical framework for the analysis of social networks of wild animal populations.