[00560] Semiparametric Kernel Estimation with Bayesian Bandwidths for Multivariate Nonnegative Data
Session Time & Room : 3E (Aug.23, 17:40-19:20) @E501
Type : Contributed Talk
Abstract : We introduce a flexible semiparametric kernel method for smoothing distributions on nonnegative supports. This multivariate estimator is guided by a given parametric part, here an uncorrelated exponential distribution estimated by maximum likelihood, and a nonparametric part which is a weight function to be smoothed through multiple gamma kernels. Also, a diagnostic model discusses the choice between the parametric, semiparametric and nonparametric approaches. Finally, practical multivariate semicontinuous datasets illustrate the usefulness of the method.