A Nonparametric Change-Point Control Chart

Abstract

The assumption of fully known in-control distributions has long been recognized as an idealization, at best approximately true. Recent development of normal-based change-point methods has allowed the assumption of exactly known in-control mean and variance to be relaxed, but retained the assumption of normality. In this paper, we develop a nonparametric tool based on the change-point model for statistical process control. This method is shown to perform well, even beating the parametric approach for small to moderate shifts in normal data, and to involve relatively light computation.