Prediction Analysis of Cross Classifications.

evaluate an author's presentation, analysis and interpretation of data." (p. vii) This purpose is accomplished in any easy-to-read introductory text for the medical student or physician, which, using a minimum of mathematical development, illustrates the many uses of statistics in medicine. The major strengths of this book may be found in Colton's use of the richness of the medical literature to provide numerous carefully and completely worked examples and exercises with answers; his attention throughout the book to concepts and methods relevant to medical research, such as vital statistics, rates, life tables, clinical trials (with discussions of randomization and historical controls), and medical surveys; his unique discussion in part III of fallacies in numerical reasoning; and his guide to critical reading of medical literature. The chapter on fallacies provides examples of incorrect reasoning iesulting from, e.g., lack of denominator data in compalisons of groups, improper choice of comparison groups, and Berkson's fallacy. The coverage of topics in the three sections of this book is adequate for its intended audience. "Part I: Fundamentals" includes descriptive statistics and an elementary introduction to probability (including an illustration of the use of the normal distribution in a blood donor screening problem); "Part II: Statistical Inference" includes inference on means and proportions, regression and correlation with brief mention of topics beyond linear regression, nonparametric methods, sequential analysis, and longitudinal data and use of life tables; "Part III: Statistics in Medical Research," is discussed above. Some minor annoyances, at least to this reviewer, were the difficulty in locating definitions which are interspersed in the text and the use of nonstandard notation such as ir for the binomial probability of success. Also, the definition of incidence rate on page 46, giving as its denominator "the population in risk but free of the disease at the beginning of the time interval," holds only in a closed population. For all practical purposes, the denominator would be defined as the average population at risk during the time interval. The value of this book would have been enhanced if the author had discussed even more of the methods used in medical research, such as comparison of survival in subgroups. Colton wisely pays considerable attention to the discussion of the assumptions of the methods discussed as well as interpretation of results. Statistics in Medicine is a useful elementary text which gives careful attention to many of the statistical methods and issues medical students and physicians are confronted with in the current medical literature. JUDITH D. GOLDBERG Mt. Sinai School of Medicine