Causal inference is a complex scientific task that relies on evidence from multiple sources and a variety of methodological approaches. By providing a cohesive presentation of concepts and methods that are currently scattered across journals in several disciplines, Causal Inference: What If provides an introduction to causal inference for scientists who design studies and analyze data. The book is divided into three parts of increasing difficulty: causal inference without models, causal inference with models, and causal inference from complex longitudinal data.
FEATURES:
* Emphasizes taking the causal question seriously enough to articulate it with sufficient precision
* Shows that causal inference from observational data relies on subject-matter knowledge and therefore cannot be reduced to a collection of recipes for data analysis
* Describes causal diagrams, both directed acyclic graphs and single-world intervention graphs
* Explains various data analysis approaches to estimate causal effects from individual-level data, including the g-formula, inverse probability weighting, g-estimation, instrumental variable estimation, outcome regression, and propensity score adjustment
* Includes software and real data examples, as well as 'Fine Points' and 'Technical Points' throughout to elaborate on certain key topics
Causal Inference: What If has been written for all scientists that make causal inferences, including epidemiologists, statisticians, psychologists, economists, sociologists, political scientists, computer scientists, and more. The book is substantially class-tested, as it has been used in dozens of universities to teach courses on causal inference at graduate and advanced undergraduate level.
Rezensionen / Stimmen
"With this clear rigorous, and readable presentation of models for causal inference using potential outcomes and counterfactuals, Hernan and Robins have provided a text that will be useful and enjoyable for students, practitioners, and researchers in statistics and applied fields."
- Andrew Gelman, Columbia University, USA
"This is the definitive book on modeling causal effects and conducting statistical inference with the resulting models, invaluable both as teaching and reference resource. It brings together a vast range of developments over recent decades in a well-organized manner, with exceptionally clear descriptions of the models, methods, emphasizing their motivations in scientific questions and goals. Especially valuable are the careful links drawn from everyday notions of causal and selection effects to modern causal models, including models for estimating and optimizing effects of chronic treatments - those administered over sustained periods of time, as are now standard in modern preventive medicine."
- Sander Greenland, Emeritus Professor of Epidemiology and Statistics, University of California Los Angeles
"This is the most eagerly anticipated book ever in its field. It provides a lean, readable, comprehensive and coherent formulation of methods for strengthening causal inference in (largely) non-experimental data. In crystal clear language it takes the reader through the logic and practice of setting up and analysing a causal question, and, if taken literally, should reveal the many situations when the data available to the investigator do not provide an adequate basis for asking specific causal questions. It provides the ideal core text for a course on causal inference that will be of relevance to many disciplines. The book has been carefully hewn over a decade through public scrutiny, response and revision, with drafts having been made available on-line and, in addition to the fabulous content, it will surely be seen as a truly innovative model for future book production. Because of its unique pre-publication history it has become a citation classic over the last 10 years. I can think of few epidemiologists who will not want to own their own copy of this book, and this will likely apply to investigators in many other fields."
- George Davey Smith FRS, Director of the Medical Research Council Integrative Epidemiology Unit (IEU), University of Bristol, UK
"Hernan and Robins have contributed enormously to the concepts and methods of causal inference. Now their unique collaboration has culminated in an invaluable gift to any researcher in the biomedical or social sciences, a Rosetta Stone for causal inference. The ideas are gathered from myriad sources in diverse disciplines, and woven into one coherent package of logical exposition. In part I, Causal Inference Without Models, they elucidate how to frame a causal question so that it is answerable, while spotlighting the assumptions needed to reach a correct inference. In part II, they progress to using parametric modeling, and in part III they explore the intricacies of evaluating causal questions using complex longitudinal data. The language, reflecting the thinking, is direct and clear. With this work, they have performed a great service for generations of students and researchers."
- Kenneth Rothman, Boston University, USA
"Anyone interested in data science and machine learning should read this book. To form valid models of the world from data, understanding causal inference is critical. They've provided a comprehensive and accessible introduction to the topic."
- Suchi Saria, John Hopkins University, USA
Sprache
Verlagsort
Verlagsgruppe
Zielgruppe
Für Beruf und Forschung
Professional and Professional Practice & Development
Produkt-Hinweis
Illustrationen
128 s/w Abbildungen, 128 s/w Zeichnungen, 21 s/w Tabellen
21 Tables, black and white; 128 Line drawings, black and white; 128 Illustrations, black and white
Maße
Höhe: 280 mm
Breite: 210 mm
Gewicht
ISBN-13
978-1-4200-7616-5 (9781420076165)
Copyright in bibliographic data and cover images is held by Nielsen Book Services Limited or by the publishers or by their respective licensors: all rights reserved.
Schweitzer Klassifikation
Miguel Hernan conducts research to learn what works to improve human health. Together with his collaborators, he designs analyses of healthcare databases, epidemiologic studies, and randomized trials. Miguel teaches clinical epidemiology at the Harvard-MIT Division of Health Sciences and Technology, and causal inference methodology at the Harvard T.H. Chan School of Public Health, where he is the Kolokotrones Professor of Biostatistics and Epidemiology. His edX course "Causal Diagrams" is freely available online and widely used for the training of researchers.
James Robins is a world leader in the development of analytic methods for drawing causal inferences from complex observational and randomized studies with time-varying treatments. His contributions include new classes of estimators based on the g-formula, inverse probability weighting of marginal structural models, and g-estimation of structural nested models. He teaches advanced epidemiologic methods at the Harvard T.H. Chan School of Public Health, where he is the Mitchell L. and Robin LaFoley Dong Professor of Epidemiology.
Autor*in
Harvard School of Public Health, Boston, Massachusetts, USA
Harvard School of Pub Health, Boston, Massachusetts, USA
Part I: Causal inference without models 1. A definition of causal effect 2. Randomized experiments 3. Observational studies 4. Effect modification 5. Interaction 6. Graphical representation of causal effects 7. Confounding 8. Selection bias 9. Measurement bias 10. Random variability Part II: Causal inference with models 11. Why model? 12. IP weighting and marginal structural models 13. Standardization and the parametric g-formula 14. G-estimation of structural nested models 15 Outcome regression and propensity scores 16. Instrumental variable estimation 17. Causal survival analysis 18 Variable selection for causal inference Part III: Causal inference from complex longitudinal data 19. Time-varying treatments 20. Treatment-confounder feedback 21. G-methods for time-varying treatments 22. Target trial emulation 23. Causal mediation
