Abstract: We define a nonparametric prewhitening method for estimating conditional quantiles based on local linear quantile regression. We characterize the bias, variance and asymptotic normality of the proposed estimator. Under weak conditions our estimator can achieve bias reduction and have the same variance as the local linear quantile estimators. A small set of Monte Carlo simulations is carried out to illustrate the performance of our estimators. An application to US gross domestic product data demonstrates the usefulness of our methodology.
Key words and phrases: Local linear quantile regression, nonparametric quantile regression, prediction interval, prewhitening estimator, weighted Nadaraya-Watson estimator.