Random Measures from Martingales; applications to Bayesian nonparametrics

From an observed data set of size n, let P_n be an estimator of the distribution from where the sample came. The aim is to construct a sequence of random distributions, (P_N) for N>=n, which converges with probability one to a random distribution function P_infty, satisfying E [P_infty | P_n ] = P_n. The relevance of the construction of such sequences will be the focus of the presentation as will a description of how to construct such a sequence. The talk is based on the paper "Martingale posterior distributions" by Fong, Holmes and Walker.