Statistical methods for testing, development and manufacturing, Forrest W. Breyfogle III. Published by John Wiley & Sons Ltd, New York, 1992. ISBN 0 471 54035 8, 516 pages. Price: £52.00, Hard Cover

This book is intended to show both the manager and the industrial practitioner the power of statistical methods for producing more competitive products with a minimal effort. Its aim is to serve as an ‘easy-to-use’ manual for professionals with minimal statistical training. As a general comment, most descriptions of the methods are too sketchy, giving little explanation and no theoretical background at all. I cannot quite agree with the author that theory may only confuse the practitioner. There are many good examples included, but they cannot really make up for the lack of explanations in the general part. Each chapter starts with an overview discussion. A ‘getting started’ chapter helps the reader who does not want to study the whole book to find the chapters relevant to his or her problem. Most chapters end with some ‘do-itsmarter’ considerations. These sections summarize the material covered and give useful advice and cross-references to other chapters. There are 22 chapters in all, covering a wide range of topics including factorial experiments, reliability testing and statistical process control. The necessity of applying a combination of these techniques is strongly emphasized. Some mathematics, computer programs, computer output descriptions and various reference tables are collected in the appendices. Chapter 4, ‘Descriptive statistics and experimental traps’ covers much more than is indicated by its title; all basic concepts such as sample and population, hypothesis testing, confidence intervals, random sampling, experimental error, etc., are presented within 20 pages. Of course, there cannot be much space left for explanations! For example, it is frequently mentioned that samples have to be taken at random, leaving it to the reader to create or maintain his or her private idea of randomness. Equally, the concept of a hypothesis test remains vague. There is no word about possible misinterpretations or the imbalance between the null hypothesis and the alternative. Chapters 7 and 8 give formulae for tests and confidence intervals for means, standard deviations and proportions. All testing problems are accompanied by sample size considerations. Non-parametric tests are not included. The t-test is recommended even for obviously non-normal data. This is surprising, as the author repeatedly emphasizes the need for quick methods requiring minimal sample sizes. Chapters 10 to 19 are devoted to the main topics of the book and a wide range of material is included. This part gives a good overview of the techniques used in industrial applications, but again, it is too brief to serve as a helpful guide for the novice. In conclusion, the book describes many statistical methods which should certainly be of interest for everybody involved in software development or quality assurance and control. However, the non-statistician who wants to apply these methods will have to seek additional help; the statistician can choose among excellent books on factorial experiments, reliability theory or process control; and the manager who needs to be economical might be reluctant to spend more than f50 or too busy to read 516 pages merely to get an overview.