Spring 2002 Seminar Series - January 29, 2002
University of Minnesota
School of Statistics
College of Liberal Arts

Error density and distribution function estimation in nonparametric regression

Fuxia Cheng
Michigan State University
(Statistics Search Candidate)

Tuesday, January 29, 2002
4:00 PM, B10 Ford Hall
Minneapolis, East Bank Campus
Social at 3:30 PM, 300 Ford Hall

Abstract

Over the last five decades, the statistical literature has been replete with papers on the estimation of the regression function. Relatively little is known about the estimation of the error density and distribution functions in these models. It is often of interest and of practical importance to know the nature of the error distribution after estimating a regression function.
This talk will discuss some asymptotics of some error density and distribution function estimators in nonparametric regression models. In particular, the histogram type density estimators based on nonparametric regression residuals obtained from the full sample are shown to be uniformly almost surely consistent. A similar result with a rate is obtained for the empirical d.f. of these residuals. Furthermore, if one uses a part of the sample to estimate the regression function and the other part to estimate the error density, then the asymptotic distribution of the maximum of a suitably normalized deviation of the density estimator from the true error density function is the same as in the case of the one sample setup. Similarly, a suitably standardized nonparametric residual empirical process based on the second part of the sample is shown to weakly converge to a time transformed Brownian bridge.