Medical Statistics

This book is suitable to be used as a textbook for all levels of students in medical school. It is also useful as a reference book for students interested in the application of biostatistics in medicine. Materials from the Introduction to Chapter 6 are similar to those of an elementary statistical textbook.


This book is more modern than the current textbook in medical statistics. In this book, biostatistics and epidemiologic concepts are nicely blended. In contrast to the fallacy of the p-value, it introduces the Bayes factor as a measure of the evidence hidden in the sample data. It illustrates the application of the regression to the mean in medicine. Many epidemiologic concepts such as sensitivity and specificity of the diagnostic test, classification and discrimination, types of bias, etc. are discussed in the book.


Chapter 7 on "Correlation and Regression" includes the concept of regression to the mean, generalized linear (Poisson and Logistic) regression models, and discrimination of new data to belong to which sample data sets. Chapter 8 covers the nonparametric inference, including Kolmogorov and Smirnov test. Via the estimation and hypothesis testing, sample sizes are determined in Chapter 9. Chapter 10 discusses the study of design for collecting sample data, including cohort, cross-sectional, case-control, and clinical trial. In addition, types of bias are expounded as a last section in Chapter 10.


Chapter 11 covers in detail the inference on contingency tables, including 2 x 2, two-way, and three-way. Five tests (Pearson, log-odds-ratio, Fisher-Irwin, McNemar, and Ejigou-McHugh) are listed in Section 11.1. Six tests (Pearson, First-order interaction, Yate's linear trend, Stuart's marginal homogeneity, Kendall, and Wilcoxon-Mann-Whitney) are described in Section 11.2. Three tests (Pearson, log-odds-ratio on first-order interaction, Barlett's on second-order interaction) and Simpson's paradox are covered in Section 11.3.


Chapter 12 covers analysis of survival data. Two methods (life-table and Kaplan-Meier) are introduced for estimating the survivor function in Section 12.2. Four methods (maximum likelihood, Armitage's preference, Wald's sequential sign, and Armitage's restricted sequential) for comparing two survival curves are covered in Section 12.3. Proportional hazard model and the log-rank test are discussed, respectively, in Section 12.4 and 12.5.


In addition, advanced techniques in comparing two survival curves are included in the book such as Armitage's preference method, Armitage's restricted sequential test and Wald's sequential sign test. Also, inference on contingency tables are treated in more detail than other books.