A multinomial network method for the analysis of mate choice and assortative mating in spatially structured populations

作者：Muniz Danilo G^*; Santos Eduardo S A; Guimaraes Paulo R Jr; Nakagawa Shinichi; Machado Glauco

来源：Methods in Ecology and Evolution, 2017, 8(10): 1321-1331.

DOI：10.1111/2041-210X.12798

• assortativeness
• information filtering
• missing data
• preference function
• preference strength
• sexual selection
• social networks
• spatial structure

Mate sampling, whereby individuals cannot access all potential mating partners in a population, is a ubiquitous yet poorly explored process. Ignoring mate sampling may underestimate female choice because the smaller the sample taken by individuals of the choosing sex, the weaker the correlation between sexually selected traits and the mating success among individuals of the chosen sex. A main factor promoting mate sampling is the spatial distribution of individuals. Thus, including distances in models of mate choice should improve estimates of mate choosiness. However, spatial distances between individuals are pairwise variables and cannot be readily included in the models commonly used to investigate mate choice.
We address this limitation by proposing a multinomial network (MN) model of mate choice, and comparing its performance with a previously published binomial generalized linear mixed model. Both models allow the inclusion of pairwise predictors, accommodating spatial distances between individuals in analyses of mate choice. We evaluated the performance of these models in detecting directional and assortative mate choice using different simulated datasets: with and without spatial information, and with and without spatial autocorrelation of male and female traits. We also took samples of different sizes from the simulated datasets to evaluate the models' performance when data are incomplete.
Using both models, the exclusion of spatial information underestimated mate choice. Small sample sizes from the simulated populations led to underestimated directional mate choice, whereas assortative choice estimates were unbiased. Taking larger samples increased statistical power, and confidence interval coverage of both models. Spatial autocorrelation decreased the power of both models, but the MN model was less affected by it.
We conclude that including space in analyses of mate choice increases our ability to detect and accurately estimate mate choice using observational data. The MN model is a powerful and flexible tool that should be used in studies of mate choice in spatially structured populations. Moreover, the model can be used to investigate choice in other contexts, such as floral constancy by pollinators and host plant selection by phytophagous insects.

• 出版日期2017-10