| |
| |
Preface | |
| |
| |
| |
Casual inference and observational studies | |
| |
| |
| |
An overview of methods for causal inference from observational studies | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Approaches based on causal models | |
| |
| |
| |
Canonical inference | |
| |
| |
| |
Methodologic modeling | |
| |
| |
| |
Conclusion | |
| |
| |
| |
Matching in observational studies | |
| |
| |
| |
| |
The role of matching in observational studies | |
| |
| |
| |
Why match? | |
| |
| |
| |
Two key issues: balance and structure | |
| |
| |
| |
Additional issues | |
| |
| |
| |
Estimating causal effects in nonexperimental studies | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Identifying and estimating the average treatment effect | |
| |
| |
| |
The NSWdata | |
| |
| |
| |
Propensity score estimates | |
| |
| |
| |
Conclusions | |
| |
| |
| |
Medication cost sharing and drug spending in Medicare | |
| |
| |
| |
| |
Methods | |
| |
| |
| |
Results | |
| |
| |
| |
Study limitations | |
| |
| |
| |
Conclusions and policy implications | |
| |
| |
| |
A comparison of experimental and observational data analyses | |
| |
| |
| |
| |
Experimental sample | |
| |
| |
| |
Constructed observational study | |
| |
| |
| |
Concluding remarks | |
| |
| |
| |
Fixing broken experiments using the propensity score | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
The lottery data | |
| |
| |
| |
Estimating the propensity scores | |
| |
| |
| |
Results | |
| |
| |
| |
Concluding remarks | |
| |
| |
| |
The propensity score with continuous treatments | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
The basic framework | |
| |
| |
| |
Bias removal using the GPS | |
| |
| |
| |
Estimation and inference | |
| |
| |
| |
Application: the Imbens-Rubin-Sacerdote lottery sample | |
| |
| |
| |
Conclusion | |
| |
| |
| |
Causal inference with instrumental variables | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Key assumptions for the LATE interpretation of the IV estimand | |
| |
| |
| |
Estimating causal effects with IV | |
| |
| |
| |
Some recent applications | |
| |
| |
| |
Discussion | |
| |
| |
| |
Principal stratification | |
| |
| |
| |
| |
Introduction: partially controlled studies | |
| |
| |
| |
Examples of partially controlled studies | |
| |
| |
| |
Principal stratification | |
| |
| |
| |
Estimands | |
| |
| |
| |
Assumptions | |
| |
| |
| |
Designs and polydesigns | |
| |
| |
| |
Missing data modeling | |
| |
| |
| |
Nonresponse adjustment in government statistical agencies: constraints, inferential goals, and robustness issues | |
| |
| |
| |
| |
Introduction: a wide spectrum of nonresponse adjustment efforts in government statistical agencies | |
| |
| |
| |
Constraints | |
| |
| |
| |
Complex estimand structures, inferential goals, and utility functions | |
| |
| |
| |
Robustness | |
| |
| |
| |
Closing remarks | |
| |
| |
| |
Bridging across changes in classification systems | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Multiple imputation to achieve comparability of industry and occupation codes | |
| |
| |
| |
Bridging the transition from single-race reporting to multiple-race reporting | |
| |
| |
| |
Conclusion | |
| |
| |
| |
Representing the Census undercount by multiple imputation of households | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Models | |
| |
| |
| |
Inference | |
| |
| |
| |
Simulation evaluations | |
| |
| |
| |
Conclusion | |
| |
| |
| |
Statistical disclosure techniques based on multiple imputation | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Full synthesis | |
| |
| |
| |
SMIKe and MIKe | |
| |
| |
| |
Analysis of synthetic samples | |
| |
| |
| |
An application | |
| |
| |
| |
Conclusions | |
| |
| |
| |
Designs producing balanced missing data: examples from the National Assessment of Educational Progress | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Statistical methods in NAEP | |
| |
| |
| |
Split and balanced designs for estimating population parameters | |
| |
| |
| |
Maximum likelihood estimation | |
| |
| |
| |
The role of secondary covariates | |
| |
| |
| |
Conclusions | |
| |
| |
| |
Propensity score estimation with missing data | |
| |
| |
| |
| |
Introduction | |
| |
| |
| |
Notation | |
| |
| |
| |
Applied example:March of Dimes data | |
| |
| |
| |
Conclusion and future directions | |
| |
| |
| |
Sensitivity to nonignorability in frequentist inference | |
| |
| |
| |
| |
Missing data in clinical trials | |
| |
| |
| |
Ignorability and bias | |
| |
| |
| |
A nonignorable selection model | |
| |
| |
| |
Sensitivity of the mean and variance | |
| |
| |
| |
Sensitivity of the power | |
| |
| |
| |
Sensitivity of the coverage probability | |
| |
| |
| |
An example | |
| |
| |
| |
Discussion | |
| |
| |
| |
Statistical modeling and computation | |
| |
| |
| |
Statistical modeling | |