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.

Miguel Hernán 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.

Causal inference is a complex scientific task that relies on combining evidence from multiple sources, and on the application of a variety of methodological approaches. Causal Inference: What If is an introduction to causal inference when data are collected on each individual in a population. 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. The book helps scientists to generate and analyze data for causal inferences that are explicit about both the causal question and the assumptions underlying the data analysis.

• Provides a cohesive presentation of concepts and methods for causal inference that are currently scattered across journals in several disciplines
• Emphasizes the need to take the causal question seriously enough to articulate it with sufficient precision
• Shows that causal inference from observational data cannot be reduced to a collection of recipes for data analysis, as subject-matter knowledge is required to justify the necessary assumptions
• Describes causal diagrams, both directed acyclic graphs and single-world intervention graphs, to represent causal inference problems
• Describes various data analysis approaches to estimate the causal effect of interest, including the g-formula, inverse probability weighting, g-estimation, instrumental variable estimation, and propensity score adjustment
• Includes 'Fine Points' and 'Technical Points' throughout to elaborate on certain key topics, as well as software and real data examples

Causal Inference: What If has been written to be accessible to all professionals that make causal inferences, including epidemiologists, statisticians, psychologists, economists, sociologists, political scientists, computer scientists, and more. It can be used to teach an introductory course on causal inference at graduate and advanced undergraduate level.

Part I Causal Inference without Models

1. A Definition of Causal Effect
2. Randomized Experiments
3. Pbservational 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