Inference of sexual selection from pairwise contests

Sexual selection, whether mediated by male-male competition, female choice, male choice, or other factors influencing reproductive outcomes, can be usefully framed as a series of pairwise contests. Experimental evolutionary biologists have long used male-choice and female-choice experiments to quantify genotype-specific mating success, a key component of fitness. Here, we consider Drosophila mate-choice experiments and apply the Bradley-Terry model, which has been explicitly designed to quantify the merits of contestants and establish a ranking based on results from pairwise contests. The Bradley-Terry model can also be used to predict the probabilities of outcomes for contests that have not yet taken place. Leave-one-out validation allows us to assess both model fit and predictive performance for unobserved contests. After applying these methods to Drosophila mate-choice experiments, we interpret the results in terms of sexual selection and discuss the implications of deviations from a well-fitting model.