On nonparametric regression estimators based on regression quantiles

Abstract
In the ciassical regression model Yi=h(xi) + ∊ i, i=1,…,n, Cheng (1984) introduced linear combinations of regression quantiles as a new class of estimators for the unknown regression function h(x). The asymptotic properties studied in Cheng (1984) are reconsidered. We obtain a sharper scrong consistency rate and we improve on the conditions for asymptotic normality by proving a new result on the remainder term in the Bahadur representation for regression quantiles.

This publication has 3 references indexed in Scilit: