Activity Number:
|
154
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 12, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Physical & Engineering Sciences*
|
Abstract - #300573 |
Title:
|
A Diagnostic Tool for Traditional Mean and Range Control Charts
|
Author(s):
|
Cali Davis*+ and B. Adams
|
Affiliation(s):
|
University of Alabama and University of Alabama
|
Address:
|
Box 870226, Tuscaloosa, Alabama, 35487-0226, USA
|
Keywords:
|
control charting ; outliers ; contamination
|
Abstract:
|
The problem of outliers in process data is investigated. An outlier is defined to be an observation that is not representative of the true process level. Control chart signals for process data may result from process shifts or measurement problems (outliers). A diagnostic statistic is proposed as an addition to X-bar and R control charting practices. The diagnostic statistic technique is designed to delineate data problems and process problems. Such a technique would allow the use of more powerful parametric control charts if a signal occurs and no data problems are suspected. If data that produce a control chart signal are diagnosed as containing potential data problems, the process operator is deferred to robust approaches to process monitoring. Hence, the proposed method may be viewed as a hybrid of traditional and robust control charting techniques.
|