$L^1$ Bandwidth Selection in Kernel Regression Function Estimation

  • Jhun, Myong-Shic (Department of Statistics, University of Michigan Ann Arbor, MI 48105, USA)
  • Published : 1988.06.01

Abstract

Kernel estimates of an unknown regression function are studied. Bandwidth selection rule minimizing integrated absolute error loss function is considered. Under some reasonable assumptions, it is shown that the optimal bandwidth is unique and can be computed by using bisection algorithm. Adaptive bandwidth selection rule is proposed.

Keywords