Stationary distribution approximations of Two-island Wright-Fisher and seed-bank models using Stein's method

TL;DR

This paper develops explicit bounds for approximating the stationary distributions of two finite population Markov chain models, the two-island Wright-Fisher and seed-bank models, using Stein's method, depending on mutation and migration rates.

Contribution

It introduces new Stein's method-based theorems with explicit bounds for stationary distribution approximations of complex population models.

Findings

Bounds depend on mutation and migration rates

Diffusion model approximates when mutation and migration are balanced

Beta distribution approximates when migration dominates mutation

Abstract

We consider two finite population Markov chain models, the two-island Wright-Fisher model with mutation, and the seed-bank model with mutation. Despite the relatively simple descriptions of the two processes, the the exact form of their stationary distributions is in general intractable. For each of the two models we provide two approximation theorems with explicit upper bounds on the distance between the stationary distributions of the finite population Markov chains, and either the stationary distribution of a two-island diffusion model, or the beta distribution. We show that the order of the bounds, and correspondingly the appropriate choice of approximation, depends upon the relative sizes of mutation and migration. In the case where migration and mutation are of the same order, the suitable approximation is the two-island diffusion model, and if migration dominates mutation, then the weighted average of both islands is well approximated by a beta random variable. Our results are derived from a new development of Stein's method for the stationary distribution of the two-island diffusion model for the weak migration results, and utilising the existing framework for Stein's method for the Dirichlet distribution.