A Unifying Model for the Analysis of Phenotypic, Genetic, and Geographic Data

Recognition of evolutionary units (species, populations) requires integrating several kinds of data, such as genetic or phenotypic markers or spatial information in order to get a comprehensive view concerning the differentiation of the units. We propose a statistical model with a double original advantage: (i) it incorporates information about the spatial distribution of the samples, with the aim to increase inference power and to relate more explicitly observed patterns to geography and (ii) it allows one to analyze genetic and phenotypic data within a unified model and inference framework, thus opening the way to robust comparisons between markers and possibly combined analyses. We show from simulated data as well as real data that our method estimates parameters accurately and is an improvement over alternative approaches in many situations. The power of this method is exemplified using an intricate case of inter-and intraspecies differentiation based on an original data set of georeferenced genetic and morphometric markers obtained on Myodes voles from Sweden. A computer program is made available as an extension of the R package Geneland. [Bayesian model; bio-geography;