Cumulative Sum Charts and Charting for Quality Improvement

Rating:
Summary: useful introduction to CUSUM Charts
Review: Much like Shewhart charts CUSUM Charts are used to detect changes in a process over time. They are commonly used by engineers to do statistical process control. Doug Hawkins is a statistician and professor with expertise in multivariate analysis and outlier detection. Many years ago he wrote a nice little monograph on outliers. This is a nicely written text that is amenable to engineers. It introduces the reader to the basic concepts in quality control and explains in a elementary way the basic theory underlying the Shewhart and the CUSUM charts. This is followed by a chapter on design issues for CUSUM charts. Later chapters provided background on important related probability distributions, first continuous and then discrete. The general theory is given in Chapter 6. It includes some treatment of the general exponential family. Although probability theory is important, only essential topics are covered. It includes some real examples and special topics. The special topics include multivariate charting, regression adjustment, correlated data and a brief discussion of time series methods. CUSUM charts can also be used to detect outliers but are somewhat slow at detecting them. In fact, I once wrote a paper comparing CUSUM charts with a time series method for detecting outliers. However, in the control chart context the intent is to detect changes in a process which are often reflected in trends. For that purpose an outlier can lead to a false alarm signal from a CUSUM chart(very short term problem that is corrected without intervention). In Section 9.1 a Winsorization technique is describe that modifies the control chart to detect the changes in the mean or variance of the process as it is designed to but without causing the false alarm presented by a single outlier.

Rating:
Summary: Excellent source for practicing engineers
Review: Text treats both practical applications as well as limitations of the CUSUM techniques.

Underlying therory for CUSUM chart development using various statistical distributions is treated in later chapters.

The authors have included internet address for downloading application software. The software is compatable with any Windows based system and Excell package.