The Discrepancy Principle for Choosing Bandwidths in Kernel Density Estimation

TL;DR

This paper examines the discrepancy principle for selecting bandwidths in kernel density estimation, providing new theoretical insights and comparing its performance to standard methods through simulations.

Contribution

It offers new theoretical results on the consistency of the discrepancy principle and extends previous analyses under various smoothness conditions.

Findings

Discrepancy principle can be consistent under weak conditions.

Some methods perform well across different densities and sample sizes.

Performance differs from asymptotic predictions at finite sample sizes.

Abstract

We investigate the discrepancy principle for choosing smoothing parameters for kernel density estimation. The method is based on the distance between the empirical and estimated distribution functions. We prove some new positive and negative results on L_1-consistency of kernel estimators with bandwidths chosen using the discrepancy principle. Consistency crucially depends on a rather weak H\"older condition on the distribution function. We also unify and extend previous results on the behaviour of the chosen bandwidth under more strict smoothness assumptions. Furthermore, we compare the discrepancy principle to standard methods in a simulation study. Surprisingly, some of the proposals work reasonably well over a large set of different densities and sample sizes, and the performance of the methods at least up to n=2500 can be quite different from their asymptotic behavior.