A Bayesian approach to the identification of panmictic populations and the assignment of individuals.

摘要: We present likelihood-based methods for assigning the individuals in a sample to source populations, on basis of their genotypes at co-dominant marker loci. The populations are assumed be Hardy-Weinberg and linkage equilibrium, but allelic composition these even number represented treated as uncertain. parameter interest is partition set sampled individuals, induced by assignment populations. maximum likelihood method, then more powerful Bayesian approach estimating this partition. In general, it will not feasible evaluate evidence supporting each possible sample. Furthermore, when large, may supporting, individually, most plausible partitions because there many which difficult assign. To overcome problems, we use low-dimensional marginals (the 'co-assignment probabilities') posterior distribution measures 'similarity', apply hierarchical clustering algorithm identify clusters whose together well supported distribution. A binary tree provides visual representation how supports cluster hierarchy. These applicable other problems where set. Because co-assignment probabilities independent arbitrary labelling avoid label-switching problem previous methods.