Strengths: No mathemetical formulas, minimal statistical jargon, and good use of clinical data and graphs
Weaknesses: Lack of formal definitions and formulas, no discussion of computer applications
Audience: Students and practitioners in the health sciences who have no previous exposure to statistics
It starts with a discussion of the types of data usually involved in health sciences, with examples from clinical situations. In discussing quantitative data, the author should have included interval scale data with no natural origin, such as intelligence-quotient scores. Reader `s should have been informed that interval scale variables are often used in health assessment and may require different statistical analysis.
Basic statistical concepts and terminology are introduced in the second chapter on summarizing univariate data. Various types of data, basic descriptive statistics such as the mean, median and range, and statistical concepts such as the normal distribution are introduced through the clever use of graphics.
In chapter 3, methods for summarizing data with two or more variables are discussed, starting with various combinations of data types in bivariate data sets. I was very impressed with this chapter, and I am sure that the audience will pick up the concepts without difficulty; however, the chapter does not provide any practical means to handle actual data or to review others' work critically. With the use of 2 x 2 contingency tables, the basic concepts of probability are introduced. Sensitivity, specificity and predictive values are also discussed in this contxAlthough the author tries to emphasize the importance of predictive values in individual case management and their dependence on the population prevalence, he should have clarified that the predictive values depend both on the sensitivity of the test used and the prevalence in the risk population. Furthermore, it would have been beneficial for readers if he had extended his discussion to medical decision making and some of its cost aspects.
The basic logic of hypothesis testing and the rationale for confidence intervals are presented in chapters 5 and 6. The author's definition and interpretation of p values are appropriate, but there is some danger of misuse if readers follow the author's description. Readers should have been warned against adjusting the significance level after the computation of the p value, because of the danger of manipulating research outcomes. The last chapter is devoted to sampling and statisti-cal power. Type I and II errors are discussed, and various sampling schemes are presented. The presentation in this chapter is not as clear or simple as in other chapters. The author may have overextended the scope and intention of this book: sample size determination and sampling methods should have been left to the experts.
An unusual characteristic of the approach in this book is that no mathematical formulas are used and no formal definitions for the statistical terms are presented. This approach will be welcomed by those who have an aversion to mathematics and statistical jargon; however, after reading the book, readers will find it difficult to locate formal definitions or formulas for resolving daily problems. Thus, the book is an excellent starting point for those who are not mathematically oriented, but no more. The book has very little value as a reference on readers' bookshelves. An appendix with a list of statistical formulas and definitions, with a proper index to the main text, would have improved the book's usefulness.
Another shortcoming concerns computer applications. The use of a computer is an essential part of present-day statistical analysis, and some exercises to be done with the use of a computer would have enhanced the learning process. However, the book ignores this aspect of statistics. The author appears to believe that computer applications are beyond the level of the readers. However, in view of the current availability of personal computers and of many excellent easy-to-use statistical software packages, I feel that computers are indispensable in learning statistical methods.
I recommend this book as preliminary reading for medical and health science students or practitioners who have never been exposed to statistics but are interested in statistical methods. I would also suggest this book as review reading for my public health sciences students before they take a statistics and data analysis course. Alternatively, it may be used in combination with statistical software or spreadsheet applications as a textbook for a short preparatory course in statistics and data analysis.
Kyung S. Bay, PhD
Professor
Department of Public Health Sciences
Faculty of Medicine
University of Alberta
Edmonton, Alta.