Using the Minimized Integrated Square Error Estimator for Phase I Control Charting

• Activity Number: 154
• Type: Contributed
• Date/Time: Monday, August 12, 2002 : 2:00 PM to 3:50 PM
• Sponsor: Section on Quality & Productivity*
• Author(s): John Dyer*+ and Michael Conerly
• Affiliation(s): Georgia Southern University and University of Alabama
• Keywords: Maximum Likelihood Estimation ; Minimum Distance Estimation ; Phase I control charting ; Outlier Detection ; Statistical Process Control

Abstract

Statistical Process Control (SPC) uses control charts for process monitoring. Control charting is a two-stage procedure: Phase I, the retrospective phase, and Phase II, the prospective phase. In Phase I, the process parameters are estimated from the data for the control chart. Maximum likelihood estimators (MLEs) are typically used to obtain parameter estimates based on a reasonably large set of process data, yet MLEs are not always robust to outliers. It is imperative that the data gathered during Phase I are "good data"; that is, they are representative of typical process data and free from the influence of outliers. In this article, we show that the minimized integrated square error estimator (L2E) can be used as an alternative to MLEs and is a more robust estimator for estimating control charting parameters. L2E has been used for nonparametric density estimation and has recently been shown to be appropriate for obtaining parameter estimates (for continuous distributions such as the normal distribution) for large data sets that may contain outliers. An example of the design and use of the L2E criterion for control chart parameter estimation is provided.