Nonparametric Estimation of Location

Publisher Website
Google Scholar
Abstract
A sequence of asymptotically normally distributed estimators of location is presented having the property that, for any ∈ > 0, all estimators in the sequence beyond an appropriate point have asymptotic variances within ∈ of the Cramér-Rao lower bound, uniformly for all symmetric distributions in a nonparametric family constrained only by regularity conditions. The simplest estimator in this sequence is the familiar trimmed mean. The next simplest estimator examined in some detail is shown to possess good efficiency-robustness properties, both asymptotically and for small sample sizes. This estimator is much easier to compute than previously proposed estimators having similar properties, and a good nonparametric estimate of the variance of the location estimator is produced as a by-product.

This publication has 0 references indexed in Scilit: