Authors: John K. Mahaney, Richard J. Goeke, David E. Booth
Addresses: Metallurgical Consultants, Inc., 737 Hampton Ridge Drive, Akron, OH 44313-5082, USA. ' School of Business Administration, Widener University, Chester, PA 19013, USA. ' Department of Management and Information Systems, College of Business Administration, Kent State University, P.O. Box 5190, Kent, OH 44242, USA
Abstract: Statistical Process Control (SPC) is an integral component of nearly every industrial process, and proper outlier (out of control point) detection is crucial if processes are to remain in statistical control. Control charting methods are widely used in SPC and outlier detection, especially in manufacturing settings, but can also be useful for non-production business data as well. However, these methods require that the data under study be Independent and Identically Normally Distributed (IIND). Unfortunately, much of the industrial data studied is time-series, not IIND, rendering standard control charting methods inappropriate. This study used an ARMA (1,1) model with the Chen and Liu (1993) JE outlier detection technique, and found it superior to control charting in identifying the position and type of potential outliers (out of control points) in nine sets of non-production business data.
Keywords: outlier detection; control charts; ARMA (1,1); business data; statistical process control; SPC; out of control points; operational research.
International Journal of Operational Research, 2007 Vol.2 No.2, pp.115 - 134
Available online: 14 Feb 2007 *Full-text access for editors Access for subscribers Purchase this article Comment on this article