Averaging of density kernel estimators

Abstract

Averaging provides an alternative to bandwidth selection for density kernel estimation. We propose a procedure to combine linearly several kernel estimators of a density obtained from different, possibly data-driven, bandwidths. The method relies on minimizing an easily tractable approximation of the integrated square error of the combination. It provides, at a small computational cost, a final solution that improves on the initial estimators in most cases. The average estimator is proved to be asymptotically as efficient as the best possible combination (the oracle), with an error term that decreases faster than the minimax rate obtained with separated learning and validation samples. The performances are tested numerically, with results that compare favorably to other existing procedures in terms of mean integrated square errors.