Coverage Error Optimal Confidence Intervals for Local Polynomial Regression

TL;DR

This paper develops higher-order inference methods for local polynomial regression, providing Edgeworth expansions and coverage error bounds that improve confidence interval accuracy and include derivative estimation.

Contribution

It introduces uniform Edgeworth expansions and coverage error expansions for local polynomial regression, including boundary cases, and proposes inference-optimal bandwidth selectors.

Findings

Robust bias corrected intervals achieve fastest coverage error decay.

Derived higher-order expansions depend on smoothness and boundary location.

Implemented in R and Stata packages for practical use.

Abstract

This paper studies higher-order inference properties of nonparametric local polynomial regression methods under random sampling. We prove Edgeworth expansions for $t$ statistics and coverage error expansions for interval estimators that (i) hold uniformly in the data generating process, (ii) allow for the uniform kernel, and (iii) cover estimation of derivatives of the regression function. The terms of the higher-order expansions, and their associated rates as a function of the sample size and bandwidth sequence, depend on the smoothness of the population regression function, the smoothness exploited by the inference procedure, and on whether the evaluation point is in the interior or on the boundary of the support. We prove that robust bias corrected confidence intervals have the fastest coverage error decay rates in all cases, and we use our results to deliver novel, inference-optimal bandwidth selectors. The main methodological results are implemented in companion \textsf{R} and \textsf{Stata} software packages.