Browse Books

Reviewer: Curtis Roger Cook

Measurement in software engineering is a huge and multifaceted topic. Among the choices for coverage in such a book are comprehensive treatment of most or all topics; focused coverage of a few topics; and cursory treatment of most topics. The author has chosen the last, with an emphasis on the integration of measurement into the practice of software engineering. He compensates for lack of depth with many references at the end of each chapter. The book is broken into two sections: “Foundations” (chapters 1–5) and “Supporting Topics” (chapters 6–8). Chapter 1 introduces measurement theory (measurement scales and numerical relation systems) and provides motivation for and examples of measurement in software engineering. Chapter 2 concentrates on software design and its evaluation by module coupling and cohesion. Only information flow metrics and measures of module size are presented in any detail. Chapter 3 introduces quality and quality standards (primarily the ISO 9001 approach) and the use of product metrics to identify potential problem areas and to indicate actual problems. Chapter 4 is a general introduction to the use of measures in software project management. It contains sections on productivity measures (primarily function points) and cost and effort prediction models. Shepperd makes the important point that quantitative prediction models must be calibrated to the particular environment and application and that historical data must be collected in order to do the calibration. Chapter 5 presents several methods and frameworks for the collection of software measurement data. Chapter 6 introduces formal models of system architecture, with an emphasis on the algebraic approach. It presents a detailed model and discusses its use in reasoning about design measurements. Chapter 7 is a sketchy and informal introduction to statistical analysis and experimental design. Chapter 8 briefly presents process models and gives an example of integrating measurement into a process model. The reader is assumed to have some software knowledge but no experience with metrics or measurement. Although the selection of topics is adequate given the intended audience, the topic coverage tends to reflect the author's research interests, and several important and practical uses of measurements are omitted. For example, measures of the effectiveness of defect detection methods would interest many readers. Object-oriented design measures and process metrics are two topics that deserve more coverage. The topic ordering could be improved. Correlation and other statistical measures are mentioned numerous times in the first five chapters, with the reader either given a terse description of the measure or referred to chapter 7. I would make the chapter on statistics the second or third chapter to alleviate this problem. Also, the chapter on statistics is too sketchy; it just touches on several important concepts and omits others, such as the chi-squared test, which is used frequently because much of the data is nominal scale. The author does a commendable job of keeping definitions and terminology to a minimum, but in a few instances he omits definitions of terms or uses terms before defining them. The statistics chapter could stand more definitions. The examples in the chapters were for the most part well chosen. One glaring problem (at least in the copy of the book I reviewed) was the 14 totally blank pages in chapter 4, a key chapter showing the integration of measurement into project management. The book would interest anyone with some software engineering experience who wants a general overview of the use of measurement in software engineering practice. It is not a reference or a textbook.