CMStatistics 2018: Start Registration
View Submission - CMStatistics
B1591
Title: Improper priors for nonparametric Bayes estimation of Poisson intensity functions Authors:  Fumiyasu Komaki - The University of Tokyo (Japan) [presenting]
Abstract: The nonparametric Bayes estimation of intensity functions of inhomogeneous Poisson processes is investigated. It is shown that reasonable nonparametric Bayes estimators can be constructed by using improper priors, although improper priors have not been widely used for nonparametric Bayes estimation. We propose a class of improper priors for Poisson intensity functions. Nonparametric Bayes estimators based on the priors in our class are admissible under the Kullback-Leibler loss.