Preface

I have always thought of myself as something of an outsider and troublemaker, so it is with some humility that I prepare the seventh edition of this book, 30 years after the first edition appeared. Then, as now, the book had an unusual perspective: that many papers in the medical literature contained avoidable errors. At the time, the publisher, McGraw-Hill, expressed concern that this “confrontational approach” would put off readers and hurt sales. They also worried that the book was not organized like a traditional statistics text.

Time has shown that the biomedical community was ready for such an approach and the book has achieved remarkable success.

The nature of the problems with the medical literature, however, has evolved over time and this new edition reflects that evolution. Many journals now have formal statistical reviewers so the kinds of simple errors that used to dominate have been replaced with more subtle problems of biased samples and underpowered studies (although there are still more than enough inappropriate t tests to go around). Over time, this book has evolved to include more topics, such as power and sample size, more on multiple comparison procedures, relative risks and odds ratios, and survival analysis.

In this edition I actually pruned back the discussion of multiple comparison testing to focus on Bonferonni, Holm, and Holm-Sidak corrected tests for both parametric and nonparametric methods.

At the same time, this is the most extensive revision done for a new edition since the book was first published. The book is now published in a larger, more open text format with more worked out examples. There are new brief introductions to higher order analysis of variance, multiple regression and logistic regression,* as well as expanded discussions of problems with study designs and more information on how to combine information from many different studies. The examples and problems have been extensively reworked, with almost all coming from studies published in the twenty-first century.

This book has its origins in 1973, when I was a postdoctoral fellow. Many friends and colleagues came to me for advice and explanations about biostatistics. Since most of them had even less knowledge of statistics than I did, I tried to learn what I needed to help them. The need to develop quick and intuitive, yet correct, explanations of the various tests and procedures slowly evolved into a set of stock explanations and a two-hour slide show on common statistical errors in the biomedical literature and how to cope with them. The success of this slide show led many people to suggest that I expand it into an introductory book on biostatistics, which led to the first edition of Primer of Biostatistics in 1981.

As a result, this book is oriented as much to the individual reader—whether he or she is a student, postdoctoral research fellow, professor, or practitioner—as to the student attending formal lectures.

This book can be used as a text at many levels. It has been the required text for the biostatistics portion of the epidemiology and biostatistics course required of medical students, covering the material in the first eight chapters in eight one-hour lectures. The book has also been used for a more abbreviated set of lectures on biostatistics (covering the first three chapters) given to our dental students. In addition, it has served me (and others) well in a one-quarter four-unit course in which we cover the entire book in depth. This course meets for four lecture hours and has a one-hour problem session. It is attended by a wide variety of students, from undergraduates through graduate students and postdoctoral fellows, as well as faculty members.

Because this book includes the technical material covered in any introductory statistics course, it is suitable as either the primary or the supplementary text for a general undergraduate introductory statistics course (which is essentially the level at which this material is taught in medical schools), especially for a teacher seeking a way to make statistics relevant to students majoring in the life sciences.

This book differs from other introductory texts on biostatistics in several ways, and it is these differences which seem to account for the book's enduring popularity.

First, because inappropriate use of the t test to analyze multigroup studies continues to be a common error, probably because the t test is usually the first procedure presented in a statistics book that will yield the highly prized P value. Analysis of variance, if presented at all, is deferred to the end of the book to be ignored or rushed through at the end of the term. Since so much is published that probably should be analyzed with analysis of variance, and since analysis of variance is really the paradigm of all parametric statistical tests, I present it first, then discuss the t test as a special case.

Second, in keeping with the problems that I see in the literature, there is a discussion of multiple comparison testing.

Third, the book is organized around hypothesis testing and estimation of the size of treatment effects, as opposed to the more traditional (and logical from a theory of statistics perspective) organization that goes from one-sample to two-sample to general k-sample estimation and hypotheses testing procedures. This approach goes directly to the kinds of problems one most commonly encounters when reading about or doing biomedical research.

The examples are based mostly on interesting studies from the literature and are reasonably true to the original data. I have, however, taken some liberty in recreating the raw data to simplify the statistical problems (for example, making the sample sizes equal) so that I could focus on the important intuitive ideas behind the statistical procedures rather than getting involved in the algebra and arithmetic. There are still some topics common in introductory texts that I leave out or treat implicitly. There is not an explicit discussion of probability calculus and expected values and I still blur the distinction between P and α.

As with any book, there are many people who deserve thanks. Julien Hoffman gave me the first really clear and practically oriented course in biostatistics, which allowed me to stay one step ahead of the people who came to me for expert help. Over the years, Virgina Ernster, Susan Sacks, Philip Wilkinson, Marion Nestle, Mary Giammona, Bryan Slinker, Jim Lightwood, Kristina Thayer, Joaquin Barnoya, Jennifer Ibrahim, and Sara Shain helped me find good examples to use in the text and as problems. Bart Harvey and Evelyn Schlenker were particularly gracious in offering suggestions and detailed feedback on the new material in this edition. I thank them all. Finally, I thank the many others who have used the book, both as students and as teachers of biostatistics, who took the time to write me questions, comments, and suggestions on how to improve it. I have done my best to heed their advice in preparing this seventh edition.

Many of the pictures in this book are direct descendants of my original slides. In fact, as you read this book, you would do best to think of it as a slide show that has been set to print. Most people who attend my slide show leave more critical of what they read in the biomedical literature and people who have read earlier editions said that the book had a similar effect on them. Nothing could be more flattering or satisfying to me. I hope that this book will continue to make more people more critical and help improve the quality of the biomedical literature and, ultimately, the care of people.

Stanton A. Glantz

* These issues are treated in detail in a second book on the subject of multiple regression and analysis of variance, written with the same approach in Primer of Biostatistics. It is Glantz SA, Slinker BK. Primer of Applied Regression and Analysis of Variance, 2nd ed. New York: McGraw-Hill; 2001.