Y. Zhao, M. Wiper, C. Ausin
We examine the properties of bivariate random Berstein polynomials for density estimation on the unit square. In particular, we use a Bernstein-Dirichlet prior for the class of Bernstein densities. We study the convergence rate of the posterior distribution of a bivariate Bernstein polynomial model under very general conditions. We also demonstrate that when the true data generating distribution is Bernstein, we obtain a nearly parametric convergence rate. Finally, we illustrate the practical application of bivariate Bayesian Bernstein polynomial based approximations with various simulated and real data examples.
Palabras clave: Bernstein polynomials, bayesian nonparametrics, bivariate density estimation
Programado
XC1a Pósters (Estadística)
18 de abril de 2012 12:00
Salón Madrid