A robust dynamic screening system by estimation of the longitudinal data distribution

Abstract

To online monitor the longitudinal performance of processes and give early signals to processes with irregular patterns, a series of dynamic screening systems (DySS) have been proposed in the literature. Existing DySS methods are all based on estimation of the in-control (IC) mean and variance of processes with a regular longitudinal pattern. In this paper, a new DySS method is suggested, which is based on estimation of the IC distribution of processes with a regular longitudinal pattern. Based on the estimated IC distribution, a statistical process control chart is constructed for sequentially detecting any distributional shifts in a longitudinal process. The suggested control chart is relatively simple to design and implement, and it is robust to the true IC distribution. Numerical examples show that it outperforms some representative existing DySS methods. These properties make it an ideal tool for dynamic screening applications, which is demonstrated by a real-data example.