Characterizing Markov exchangeable sequences (Q1345085): Difference between revisions

In the 1920s \textit{W. E. Johnson} discovered a simple characterization of the Dirichlet family of conjugate priors for a multinomial distribution having at least three categories. The same family of prior has been extensively used in Bayesian inferences after Ferguson proposed his non- parametric method of estimation. The author extends Johnson's results to the case of Markov exchangeable sequences. The method used by the author for this goal is simple and very elegant. The result also shows that Johnson's original sufficiency condition can be extended to and used in a domain much wider than that till now considered.