A1022
Title: Uniform convergence rates for nonparametric estimators of a density function when the density has a known pole
Authors: Sorawoot Srisuma - National University of Singapore (Singapore) [presenting]
Abstract: The aim is to study the uniform convergence rates of nonparametric estimators for a probability density function and its derivatives when the density has a known pole. Such a situation arises in some structural micro econometric models, e.g., in auction, labour, and consumer search, where uniform convergence rates of density functions are important for nonparametric and semiparametric estimation. Existing uniform convergence results based on Rosenblatt's kernel estimator are derived under the assumption that the density is bounded. They are not applicable when there is a pole in the density. The pole nonparametrically is treated, and various kernel-based estimators are shown can attain any convergence rate that is slower than the optimal rate (when the density is bounded) uniformly over an appropriately expanding support under mild conditions.
