The New Statistical Analysis of Data

Statistical Analysis Simplijed is the introductory volume in the H. James Harrington Performance Improvement Series published by McGraw-Hill. The three authors have almost a century of business and quality experience between them. They have produced a highly readable explanation of statistical thinking and the seven traditional quality-control tools. The target audience is business people with little or no qualityimprovement experience who are afraid or unaccustomed to using data and numbers. This profile does not match the typical practitioner in the Technomerrics audience, so it is not likely that this book will appeal to many of them. Other volumes in the Performance Improvement Series cover statistical process control, process engineering, and reliability analysis. Statistical Analysis Simphjied admirably eliminates jargon and keeps the theory to the absolute minimum while still building a strong case for the benefits of using numbers for improvement. To appeal to as broad an audience as possible and to keep the discussion simple, the authors use many everyday experiences to support their concepts-bowling scores for control charts, baking pies for designed experiments, and moving a wide truck over a narrow bridge to explain process capability indexes. Except for one story from their work with an insurance company, which appears at the beginning of the book, the authors do not include business examples from their consulting work. This is a most unfortunate omission and passes up an ideal opportunity to build credibility with the business reader. The chapters in the book are (I) “Measure, an Objective Language for Communication”; (2) “Patterns, Insights Into the Way Things Are”: (3) “Characterization, Using One Number to Represent a Group of Numbers”; (4) “Movement, Looking at Trends Over Time”; (5) “Judgment, When to Take Action and When to Leave Things Alone”; (6) “Usefulness, How Well Your Process Meets Your Customer’s Needs”; (7) “Experimentation, Finding New Insights”; (8) “Stratification, Dividing Data Into Subsets”; (9) “Relationships, Identifying Cause and Effect”; and (10) “Implementation, Applying What You Have Learned.” Although the book strives to avoid dependence on a statistician to explain quantitative approaches, it does not remove all dependencies. It is unlikely after reading this book that one could diagnose his/her organizational improvement needs and pick the best projects. The book leaves the impression of being a textbook for a course rather than a stand-alone how-to guide. The illustrations in the book are ample and easily read, and they reinforce the concepts very well. The bibliography is skimpy, consisting of 20 entries. Seven of the entries are the authors’ works or the works of Ernst & Young, the company of which Harrington is a principal. The latest entry is from 1995. A more comprehensive bibliography would improve the book. Each chapter concludes with a mathematical brain teaser that seems intended to interject an element of fun. Unfortunately, the teasers are not tied to the concepts presented, so they do not reinforce the goals of the book. The authors have succeeded in providing correct, as well as simple, explanations of statistical process improvement. One minor point relates to the bottom of page 117. The authors explain scatterplots and write the instructions so that they could be interpreted as permitting the reader to add a regression line by hand drawing. It would be better to explicitly recommend calculating and plotting the least squares line. The authors deserve credit for mixing the discussion of specific techniques with some of the big-picture thinking the practitioner should do when pursuing improvement. Some examples are the three things needed for success-a clear picture of the desired results, an understanding of the best way to achieve them, and a healthy dose of common sense. Another example is the three simple questions statistical analysis should answer: Am I getting the results I want? Is there too much variation in the results I get’? Are the results I get stable over time? As a result of its informal and nonthreatening approach, Sraristical Analysis Simplified is an ideal first step for considering numbers and statistics as something other than mysterious, complex, and fear inducing. Unfortunately, it offers the reader no new techniques or novel applications of existing ones. In striving so hard to be simple, it forgets to point out that some quality problems are naturally complex and require a complex strategy for success. The general practitioner and reader of Technornetrics should keep this book in mind for colleagues who are new to quality. After reading it, these colleagues should be steered to magazine articles or case-study books that can provide more in-depth treatment of how statistical techniques are selected and applied.