Lines of descent and the coalescent (Q795765)

In this paper a sampling theory for Moran's model of evolution with infinitely many alleles is developed. A population consisting of a fixed number of genes per generation is considered which evolves according to Moran's model with mutation to novel types. Choosing a random sample of n genes from the population and inquiring for the origin of these with respect to the population m generations before, the following classes my be distinguished: Some genes in the sample may have descended from one gene at time m ago and are said to form an ''old'' equivalence class \(\xi_ i\), whereas other genes may have descended from mutant ancestors more recently than m ago and belong to a ''new'' class \(\eta_ j\). If \(D_ m\) old classes and \(F_ m\) new classes are observed, the sample is characterized by the partition \(R_ m=\{\xi_ 1,...,\xi_{D_ m};\eta_ 1,...,\eta_{F_ m}\}.\) The process \(R_ m(m=0,1,2,...)\) is a Markov chain for which transition probabilities as well as exact expressions for the probability distributions \(P(F_ m,R_ m| D_ m)\) and \(P(D_ m)\) are derived. The results obtained are consistent with earlier results of Ewens and Kingman. Application to the Wright-Fisher model is discussed. Moreover, from the probability distribution of the sample partition, distributions of total number of classes, class sizes, gene numbers and new alleles have been derived.