Doubly Robust Estimation
Belinda Hernández, Ilya A. Lipkovich, Michael O’Kelly · 2014
Doubly robust (DR) estimators augment the inverse probability weighting (IPW) estimator by making use subjects with fully observed data and with partially observed data, which are used to predict (or impute) missing outcome values. DR estimators use three models: one, which models the probability of being observed (the missingness model); another, which estimates missing values by modeling the relationship between the observed data and the response (the imputation model); and an analysis model, which models the overall scientific question of interest. This chapter discusses the concept of IPW estimators before going on to describe DR estimation. It talks about the assumptions, advantages and limitations of each of these methods in turn. The chapter focuses on a particular implementation of DR estimation proposed by Vansteelandt et al. (2010) before finally providing example code showing how to implement this method in SAS.
Cited by 14 works. doi:10.1002/9781118762516.ch8
Authors
References
- An Introduction to the Bootstrap
- Estimation of Regression Coefficients When Some Regressors Are Not Always Observed
- Missing Data in Clinical Studies
- Demystifying Double Robustness: A Comparison of Alternative Strategies for Estimating a Population Mean from Incomplete Data
- Demystifying Double Robustness: A Comparison of Alternative Strategies for Estimating a Population Mean from Incomplete Data (with discussion)
- A Comparison of Multiple Imputation and Doubly Robust Estimation for Analyses with Missing Data
- Analysis of Incomplete Data Using Inverse Probability Weighting and Doubly Robust Estimators
- Doubly Robust and Multiple-Imputation-Based Generalized Estimating Equations
- Relationships of Cigarette Smoking and Alcohol Consumption to Metabolic Syndrome in Japanese Men
- A method for increasing the robustness of multiple imputation