Sampling Methods of Neutral to the Right Processes

Abstract

Since Ferguson's seminal article on the Dirichlet process, the area of Bayesian nonparametric statistics has seen development of many flexible prior classes. At the center of the development lies the neutral to the right (NTR) process proposed by Doksum. Although the class of NTR processes is very rich in its members and has well-developed theoretical properties, its application has been restricted to very small portions of the class—mainly the Dirichlet, gamma, and beta processes. We believe that this is due to the lack of flexible computational algorithms that can be used as a component in a Markov chain Monte Carlo (MCMC) algorithm.The main purpose of this article is to introduce a collection of algorithms (or a tool box), some already available in the literature and others newly proposed here, so that one can construct a suitable combination of algorithms from this collection to solve one's problem.