|
Abstract:
|
Reliable and precise risk prediction is fundamental for successful management of clinical conditions, but is also a difficult task especially when the outcome of interest is time to a rare event and the number of candidate predictors is not very small. Risk estimation procedures such as fitting time specific generalized linear model via inverse probability weighting is robust to model mis-specification, but may be inefficient in the rare event setting. Most existing methods aimed to improve the efficiency of such estimation do not perform well in the rare event setting, especially when the number of predictors is not small.In this paper, we propose a two-step imputation based augmentation procedure that can lead to much improvement in estimation efficiency and is robust to model mis-specification. We also develop regularized augmentation procedures to incorporate high dimensional covariates and procedures to improve the estimation of individualized treatment effect in risk reduction. Numerical studies suggest that outperform existing methods in efficiency gain. The proposed methods are also applied to an AIDS clinical trial for treating HIV infected patients.
|
