Principles of medical statistics (Q2724109)

This is a 700 pages strong introduction to the concepts of applied statistics in the medical sciences written by a medical doctor with a mathematical education background. The author presents a comprehensive overview supplemented by refreshing discussions which are sometimes provoking and controversially to common practice in applied statistics. In agreement with the ICH guidelines for Good Clinical Practice, preference is given to quantitative statements on the effect size like confidence intervals. Qualitative statements like tests are of minor interest as the quantification by the related p-value is doubtful and may lead to erroneous interpretations. Recommended data presentations are opposed to misleading visual arrangements which can be commonly found in the scientific literature.NEWLINENEWLINENEWLINEThe book starts with two introductory chapters what statistics is and how to express data (1. Introduction; 2. Formation, Expression, and Coding of Data). The remainder of the text is organized in four major parts of approximately equal size: I. Evaluating a Single Group of Data; II. Comparing Two Groups of Data; III. Evaluating Associations; IV. Additional Activities. The first part deals with measures of location (3. Central Index of a Group), measures of scale (4. Indexes of Inner Location; 5. Inner Zones and Spreads), and some elementary probabilistic background (6. Probabilities and Standardized Indexes). Moreover, the concept of confidence intervals is introduced (7. Confidence Intervals and Stability: Means and Medians; 8. Confidence Intervals and Stability: Binary Proportions), and resampling methods like bootstrap and jackknife are opposed to the standard parametric approach. The one-sample t-test is included as the counterpart to the parametric normal confidence region. This part is completed by visual arrangements for one-dimensional data (9. Communication and Display of Univariate Data).NEWLINENEWLINENEWLINEThe second part starts with a discussion on effect size (10. Quantitative Contrasts: The Magnitude of Distinctions) before the concept of statistical tests is introduced (11. Testing Stochastic Hypotheses). In contrast to standard textbooks first permutation tests are proposed (12. Permutation Rearrangements: Fisher Exact and Pitman-Welch Test), and then the corresponding parametric, asymptotic and nonparametric tests are considered (13: Parametric Sampling: Z and t Tests; 14. Chi-Square Test and Evaluation of Two Proportions; 15. Nonparametric Rank Tests). Again, this part is completed by visual arrangements (16. Interpretations and Displays for Two-Group Contrasts; 17. Special Arrangements for Rates and Proportions).NEWLINENEWLINENEWLINEPart III is devoted to relations between two numerically measurable variables in bivariate data (18. Principles of Association). The differences in the interpretations of regression and correlation are elaborated and common abuses are denounced (19. Evaluating Trends). Moreover, a clear distinction is made between measures of association and measures of agreement and diagnostics (20. Evaluating Concordances; 21. Evaluating ''Conformity'' and Marker Tests). This part is supplemented by a treatment of censored data and repeated measurements (22. Survival and Longitudinal Analysis).NEWLINENEWLINENEWLINEPart IV collects a couple of topics occurring frequently in statistical applications: 23. Alternative Hypotheses and Statistical ''Power''; 24. Testing for ''Equivalence''; 25. Multiple Stochastic Testing; 26. Stratifications, Matchings, and ''Adjustments''; 27. Indexes of Categorial Association; 28. Non-Targeted Analysis; 29. Analysis of Variance. This covers, besides others, sequential and interim analysis, meta-analysis, design issues, trends in contingency tables, principal components, factor and cluster analysis. At the very end, analysis of variance methods are criticized by the author to be outdated and regression is proposed instead. This opinion, however, is not supported by the examples given since they are far from being convincing.NEWLINENEWLINENEWLINEAll chapters are supplemented by a large number of examples and exercises. No mathematical rigor is claimed. Occasional proofs are deferred to an appendix. The book is augmented by an extensive bibliography of 25 pages and a useful subject index. The notions used by the author are frequently different from common practice. Though this choice of notions seems appealing in many instances it causes difficulties to relate the text to other publications. Nevertheless, this book is a helpful source for a critical review of statistical methods and a valuable reference for the criticism of bad or inappropriate presentations of data.