B ayesian Product Partition Models
Fernando Andrés Quintana, Rosângela H. Loschi, Garritt L. Page · Wiley StatsRef: Statistics Reference Online · 2018
Abstract Product partition models are a class of probability distributions whose support is the set of all possible partitions of a finite collection of experimental units. For Bayesian inference, this class of distributions can be used as a prior distribution on partitions whose prominent feature is its product form. We review how this class of probability distributions can be employed in modeling from a Bayesian perspective. Both exchangeable and nonexchangeable models are considered as well as a number of extensions that facilitate incorporating covariate dependence in the prior.