Prediction of group patterns in social mammals based on a coalescent model

This study describes a statistical model which assumes that mammal group patterns match with groups of genetic relatives. Given a fixed sample size, recursive algorithms for the exact computation of the probability distribution of the number of groups are provided. The recursive algorithms are then incorporated into a statistical likelihood framework which can be used to detect and quantify departure from the null-model by estimating a clustering parameter. The test is then applied to ecological data from social herbivores and carnivores. Our findings support the hypothesis that genetic relatedness is likely to predict group patterns when large mammals have few or no predators.