Dirichlet invariant processes and applications to nonparametric estimation of symmetric distribution functions |
| |
Authors: | SR Dalal |
| |
Institution: | Department of Statistics, Rutgers University, New Brunswick, NJ 08903, U.S.A. |
| |
Abstract: | A class of random processes with invariant sample paths, that is, processes which yield (with probability one) probability distributions that are invariant under a given transformation group of interest, are introduced and their properties are studied. These processes, named Dirichlet Invariant processes, are closely related to the Dirichlet processes of Ferguson. These processes can be used as priors for Bayesian analysis of some nonparametric problems. As an application Bayes and Minimax estimates of an arbitrary distribution, symmetric about a known point, are obtained. |
| |
Keywords: | Dirichlet processes priors Bayes estimate minimax estimate |
本文献已被 ScienceDirect 等数据库收录! |
|