ABC methods can be used for Bayesian model choice between multiple models. The ABC model choice algorithm samples models from their prior probabilities and parameters from the corresponding prior distributions. Simulated data is generated and accepted if the distance between summary statistics of the simulated and observed data is below a tolerance level. As the number of simulations increases, the ABC approximation of the Bayes factor converges to the true Bayes factor. Sufficient statistics for individual models may not be sufficient for the joint model and parameters.