Markov chain Monte Carlo without likelihoods

Paul Marjoram, John Molitor, Vincent Plagnol, Simon Tavaré

Proceedings of the National Academy of Sciences · 2003 · 1.3K citations · 14 references

DOIFull text

Open access

Concepts

TL;DR

Many stochastic simulation approaches for generating observations from a posterior distribution depend on knowing a likelihood function, which is often impossible or computationally prohibitive for complex models. The authors propose a likelihood‑free Markov chain Monte Carlo method to generate posterior samples. The method constructs a Markov chain that samples from the posterior without evaluating the likelihood, and is demonstrated on ancestral inference in population genetics. The approach also enables maximum‑likelihood estimation in frequentist settings and raises several open problems.

Abstract

Many stochastic simulation approaches for generating observations from a posterior distribution depend on knowing a likelihood function. However, for many complex probability models, such likelihoods are either impossible or computationally prohibitive to obtain. Here we present a Markov chain Monte Carlo method for generating observations from a posterior distribution without the use of likelihoods. It can also be used in frequentist applications, in particular for maximum-likelihood estimation. The approach is illustrated by an example of ancestral inference in population genetics. A number of open problems are highlighted in the discussion.

References

14