Change-Point Analysis: A Powerful New Tool For Detecting Changes

Dr. Wayne A. Taylor

Change-point analysis is a powerful new tool for determining whether a change has taken place. It is capable of detecting subtle changes missed by control charts. Further, it better characterizes the changes detected by providing confidence levels and confidence intervals. When collecting online data, a change-point analysis is not a replacement for control charting. But, because a change-point analysis can provide further information, the two methods can be used in a complementary fashion. When analyzing historical data, especially when dealing with large data sets, change-point analysis is preferable to control charting. A change-point analysis is more powerful, better characterizes the changes, controls the overall error rate, is robust to outliers, is more flexible and is simpler to use. This article describes how to perform a change-point analysis and demonstrates its capabilities through a number of examples.


Table of Contents  

Introduction
Example 1: US Trade Deficit Data
Change-Point Analysis
Procedure for Performing a Change-Point Analysis
Detecting Changes in the Variation and Other Characteristics
Handling Outliers
Example 2: Complaint Data
Example 3: Part Strength
Conclusion
References