Abstract: The problem of estimating the number of change points in a sequence of independent random variables is considered in a Bayesian framework. We find that, under mild assumptions and with respect to a suitable prior distribution, the posterior mode of the number of change points converges to the true number of change points in the frequentist sense. Furthermore, the posterior mode of the locations of the change points is shown to be within Op(logn) of the true locations of the change points where n is the sample size. The prior distribution on the locations of the change points may be taken to be uniform. Finally, some simulated results are given, showing that the method works well in estimating the number of change points.
Key words and phrases: Change points, posterior distribution.