A CROSS-VALIDATORY CHOICE OF SMOOTHING PARAMETER IN ADAPTIVE LOCATION ESTIMATION

This article proposes a new data-driven method for selecting the smoothing parameter involved in constructing kernel-based adaptive location estimators. The method consists of minimizing a cross-validatory criterion with respect to the bandwidth occurring in the kernel-type estimators of the efficient score function. It is shown that the location estimator with a data-driven bandwidth selector is indeed an adaptive estimator. A simulation study reveals that the method is also practicable, showing that our estimator performs well in comparison with some other well-known location estimators. It also shows that our method has comparable finite sample performance with the bootstrap method of selecting the smoothing parameter and yet has great computational advantages.