Название: Bayesian Thinking in Biostatistics
Автор: Purushottam W. Laud, Wesley O. Johnson, Gary L Rosner
Издательство: CRC Press
Год: 2021
Страниц: 622
Язык: английский
Формат: pdf (true)
Размер: 14.6 MB

With a focus on incorporating sensible prior distributions and discussions on many recent developments in Bayesian methodologies, Bayesian Thinking in Biostatistics considers statistical issues in biomedical research. The book emphasizes greater collaboration between biostatisticians and biomedical researchers. The text includes an overview of Bayesian statistics, a discussion of many of the methods biostatisticians frequently use, such as rates and proportions, regression models, clinical trial design, and methods for evaluating diagnostic tests.

Bayesian statistical methods have found increased use and acceptance in biomedical research. More and more clinical trials are including Bayesian considerations in their designs. Interim analyses based on posterior or predictive probabilities have appeared in protocols for clinical studies sponsored by the largest pharmaceutical companies and at internationally renowned academic medical centers. Governmental regulatory agencies, such as the U.S. Food and Drug Administration and the European Medicines Agency, are becoming more open to considering clinical trials with Bayesian designs. In our roles as biostatisticians at active academic medical centers, we have used Bayesian methods in our collaborations. We receive frequent requests for information about Bayesian methods from our colleagues in other disciplines. We teach courses on Bayesian methods 1 focusing on biostatistical applications and our students come from many different departments. Given this widespread interest-especially from our colleagues in disciplines outside of statistics-we felt the time is right for an introductory textbook that reviews Bayesian inference from the standpoint of someone whose interest is biostatistical but whose background is not.

Aside from providing a summary of the concepts and theory of Bayesian inference, information that one can find in many available texts, we also discuss methods that biomedical investigators would want to use for their research. We cover Bayesian methods for analyzing time-to-event data (i.e., survival analysis), clinical trial design, longitudinal data analysis, and diagnostic tests.

Another factor that has contributed to broader use of Bayesian methods in biostatistics is the availability of computer programs to carry out the necessary computations. The stand-alone packages WinBUGS (and its current incarnation OpenBUGS), JAGS, and Stan, as well as packages and functions written for the statistical computing environment R, have put the ability to carry out rather sophisticated and complex data analyses in the hands of more people than ever before. The availability of these tools provides the opportunity to promote the use of Bayesian biostatistical methods to a wider audience. In the book and an accompanying website, we provide programs in the BUGS language, with variants for JAGS and Stan, that one can use or adapt for one's own research.

Key Features:

Applies a Bayesian perspective to applications in biomedical science
Highlights advances in clinical trial design
Goes beyond standard statistical models in the book by introducing Bayesian nonparametric methods and illustrating their uses in data analysis
Emphasizes estimation of biomedically relevant quantities and assessment of the uncertainty in this estimation
Provides programs in the BUGS language, with variants for JAGS and Stan, that one can use or adapt for one's own research

The intended audience includes graduate students in biostatistics, epidemiology, and biomedical researchers, in general.

"This thoroughly modern Bayesian book …is a 'must have' as a textbook or a reference volume. Rosner, Laud and Johnson make the case for Bayesian approaches by melding clear exposition on methodology with serious attention to a broad array of illuminating applications. These are activated by excellent coverage of computing methods and provision of code. Their content on model assessment, robustness, data-analytic approaches and predictive assessments…are essential to valid practice. The numerous exercises and professional advice make the book ideal as a text for an intermediate-level course…" - Thomas Louis, Johns Hopkins University

"The book introduces all the important topics that one would usually cover in a beginning graduate level class on Bayesian biostatistics. The careful introduction of the Bayesian viewpoint and the mechanics of implementing Bayesian inference in the early chapters makes it a complete self- contained introduction to Bayesian inference for biomedical problems….Another great feature for using this book as a textbook is the inclusion of extensive problem sets, going well beyond construed and simple problems. Many exercises consider real data and studies, providing very useful examples in addition to serving as problems." - Peter Mueller, University of Texas

Table of Contents:

1. Scientific Data Analysis 2. Fundamentals I: Bayes Theorem, Knowledge Distributions, Prediction 3. Fundamentals II: Models for Exchangeable Observations 4. Computational Methods for Bayesian Analysis 5. Comparing Populations 6. Specifying Prior Distributions 7. Linear Regression 8. Binary Response Regression 9. Poisson and Non-linear Regression 10. Model Assessment 11Survival Modeling I: Models for Exchangeable Observations 12. Survival Modeling 2: Time-to-Event Regression Models 13. Clinical Trial Designs 14. Hierarchical Models and Longitudinal Data 15. Diagnostic Tests