Résumé
With over 60 years of applied experience, Fay and Brittain present hypothesis testing and compatible confidence intervals, emphasize strategies to address the reproducibility crisis, and provide methods for proper causal interpretation in scientific research. The book presents a full scope of tools and advice on their appropriate use in practice.Fay and Brittain present statistical hypothesis testing and compatible confidence intervals, focusing on application and proper interpretation. The emphasis is on equipping applied statisticians with enough tools - and advice on choosing among them - to find reasonable methods for almost any problem and enough theory to tackle new problems by modifying existing methods. After covering the basic mathematical theory and scientific principles, tests and confidence intervals are developed for specific types of data. Essential methods for applications are covered, such as general procedures for creating tests (e.g., likelihood ratio, bootstrap, permutation, testing from models), adjustments for multiple testing, clustering, stratification, causality, censoring, missing data, group sequential tests, and non-inferiority tests. New methods developed by the authors are included throughout, such as melded confidence intervals for comparing two samples and confidence intervals associated with Wilcoxon-Mann-Whitney tests and Kaplan-Meier estimates. Examples, exercises, and the R package asht support practical use.1. Introduction; 2. Theory of tests, p-values, and confidence intervals; 3. From scientific theory to statistical hypothesis test; 4. One sample studies with binary responses; 5. One sample studies with ordinal or numeric responses; 6. Paired data; 7. Two sample studies with binary responses; 8. Assumptions and hypothesis tests; 9. Two sample studies with ordinal or numeric responses; 10. General methods for creating decision rules; 11. K-Sample studies and trend tests; 12. Clustering and stratification; 13. Multiplicity in testing; 14. Testing from models; 15. Causality; 16. Censoring; 17. Missing data; 18. Group sequential and related adaptive methods; 19. Testing fit, equivalence, and non-inferiority; 20. Power and sample size.Michael P. Fay is a Mathematical Statistician at the National Institute of Allergy and Infectious Diseases, and previously worked at the National Cancer Institute. He has served as associate editor for Biometrics, and is currently an associate editor for Clinical Trials and a Fellow of the American Statistical Association. He is a co-author on over 100 papers in statistical and medical journals and has written and maintains over a dozen R packages on CRAN. Erica H. Brittain is Deputy Branch Chief of Biostatistics Research at the National Institute of Allergy and Infectious Diseases and has well over three decades of experience as a statistician, with previous positions at FDA, National Heart, Lung, and Blood Institute, and a statistical consulting company. Her applied work at NIH and her methodological publications in statistical journals focus on innovation in clinical trial design. She frequently serves on advisory panels for FDA and NIH, and has served as Statistical Consultant for Nature journals and Associate Editor for Controlled Clinical Trials.
