Trapezoidal rule and sampling designs for the nonparametric estimation of the regression function in models with correlated errors

TL;DR

This paper introduces a trapezoidal rule-based kernel estimator for nonparametric regression with correlated errors, analyzing its asymptotic properties, optimal design, and robustness through simulations.

Contribution

It proposes a new estimator using the trapezoidal rule for correlated data, deriving its asymptotic behavior and optimal design criteria.

Findings

The new estimator has favorable asymptotic properties.

Optimal bandwidth and design improve estimation accuracy.

Simulation results compare the new method with classical estimators.

Abstract

The problem of estimating the regression function in a fixed design models with correlated observations is considered. Such observations are obtained from several experimental units, each of them forms a time series. Based on the trapezoidal rule, we propose a simple kernel estimator and we derive the asymptotic expression of its integrated mean squared error IMSE and its asymptotic normality. The problems of the optimal bandwidth and the optimal design with respect to the asymptotic IMSE are also investigated. Finally, a simulation study is conducted to study the performance of the new estimator and to compare it with the classical estimator of Gasser and M\"uller in a finite sample set. In addition, we study the robustness of the optimal design with respect to the misspecification of the autocovariance function.