English

The Rate of Convergence for Approximate Bayesian Computation

Statistics Theory 2014-07-21 v3 Statistics Theory

Abstract

Approximate Bayesian Computation (ABC) is a popular computational method for likelihood-free Bayesian inference. The term "likelihood-free" refers to problems where the likelihood is intractable to compute or estimate directly, but where it is possible to generate simulated data XX relatively easily given a candidate set of parameters θ\theta simulated from a prior distribution. Parameters which generate simulated data within some tolerance δ\delta of the observed data xx^* are regarded as plausible, and a collection of such θ\theta is used to estimate the posterior distribution θX ⁣= ⁣x\theta\,|\,X\!=\!x^*. Suitable choice of δ\delta is vital for ABC methods to return good approximations to θ\theta in reasonable computational time. While ABC methods are widely used in practice, particularly in population genetics, study of the mathematical properties of ABC estimators is still in its infancy. We prove that ABC estimates converge to the exact solution under very weak assumptions and, under slightly stronger assumptions, quantify the rate of this convergence. Our results can be used to guide the choice of the tolerance parameter δ\delta.

Keywords

Cite

@article{arxiv.1311.2038,
  title  = {The Rate of Convergence for Approximate Bayesian Computation},
  author = {Stuart Barber and Jochen Voss and Mark Webster},
  journal= {arXiv preprint arXiv:1311.2038},
  year   = {2014}
}

Comments

25 pages, 3 figures; address the distinction between fixed number of proposals and fixed number of accepted samples more explicitly

R2 v1 2026-06-22T02:03:56.267Z