Estimation and test of jump discontinuities in varying coefficient models with empirical applications

Abstract

Varying coefficient models are very important tools to explore the hidden structure between the response and its predictors. This paper focuses on estimating and diagnosing jump discontinuities in coefficient functions. A nonparametric procedure is proposed to estimate jump discontinuities based on the Nadaraya–Watson kernel smoothing and least-squares fitting, and asymptotic properties of resulting estimators are derived. Then, a jump size-based test statistic is developed for checking whether the estimated jump discontinuities are true. A computationally feasible approximation is derived for critical values of its limiting null distribution. Monte Carlo simulations are conducted to assess the finite sample performance of the proposed methodologies, and an empirical example is discussed.