Bayesian Mixtures of Triangular distributions with application to Goodness-of-Fit Testing
Judith Rousseau, Kerrie Lee Mengersen, R. McVinish · Base Institutionnelle de Recherche de l'université Paris-Dauphine (BIRD) (University Paris-Dauphine) · 2005
Two forms of mixtures of triangular distributions are considered as an alternative to the Bernstein polynomials for Bayesian nonparametric density estimation on [0, 1]. Conditions for weak and strong consistency of the posterior distribution and a rate of convergence are established. This class of priors is applied to the problem of testing a parametric family against a nonparametric alternative.