Effect of Measurement Errors on Monitoring Multivariate Process Variability

Abstract

In many real applications of quality control schemes, there exist significant measurement errors. For monitoring multivariate process variability, we discover that the power of several control charts when the quality characteristics contain measurement errors is greater than that when the quality characteristics do not contain measurement errors. Because quality characteristics without measurement errors are more informative than those with measurement errors, this phenomenon is unreasonable. In this talk, we investigate this phenomenon on many commonly used control charts for monitoring multivariate process variability, including those derived based on sample generalized variance, sum of the standardized variance for the principal component, regression-adjusted variables, and modified likelihood ratio test. An adequate control chart for monitoring multivariate process variability when there exist measurement errors will be proposed. This is a joint work with Ying Hung.