B0686
Title: A calibration method to stabilize the estimation in missing data and causal inference
Authors: Baojiang Chen - University of Texas (United States) [presenting]
Ao Yuan - Georgetown University (United States)
Jing Qin - National Institutes of Health (United States)
Abstract: Missing data are commonly available in causal inference, where the marginal means of the outcome in both the treatment and control groups are estimated. The augmented inverse weighting (AIW) estimator was commonly used to estimate the marginal mean of the outcome due to its doubly robust property. However, the AIW estimator can be severely biased if both the propensity score (PS) and the outcome regression (OR) models are misspecified. One possible reason is that the misspecification of the PS or/and OR model yields some extreme values in these models, which can have a great influence on the marginal mean estimate. We propose a calibrated augmented inverse weighting estimator for the marginal mean, which can control for these extreme values' influence, hence providing a stable marginal mean estimator. The proposed estimator also enjoys the doubly robust property. We also introduce the Box-Cox transformation in the outcome regression model to reduce the possibility of model misspecification. A smearing estimate is used to estimate the conditional mean of the outcome. Finally, we extend this method to handle high dimensional covariates in the PS and OR models. Asymptotic results are also developed. Extensive simulation studies demonstrate that the proposed method performs better than peers by providing a more stable estimate. We apply this method to an AIDS clinical trial study.
