| الوصف: |
Ph.D. ; Medical studies frequently encounter imbalanced treatment assignments, which may lead to biased estimates of treatment effects. To address this issue and obtain unbiased estimates, the application of causal inference methods becomes essential. One notable advancement in causal inference is the doubly robust estimator (DRE), which offers two opportunities for estimates to achieve consistency. By incorporating both the propensity score and outcome components, the DRE enhances the accuracy of estimating treatment effects, thereby improving the validity of medical study findings.However, the available literature on the DRE is notably lacking when it comes to time-to-event outcomes, which are frequently encountered in medical research. Moreover, the existing methods for survival analysis often have complex forms based on optimal estimating equations, and some fail to achieve double robustness in the original sense. Furthermore, these methods often rely on subjective specification of the propensity score model using a logistic model. Also, there is a lack of a DRE developed based on the Kaplan-Meier estimator to our knowledge, which is widely used in survival analysis. To address these limitations, this dissertation proposes a novel estimator and explores its various applications in survival analysis. The first contribution is the development of a semiparametric DRE that involves the Kaplan-Meier estimator and the Stute weighted empirical form, and it exhibits a simpler summation structure compared to existing methods. Moreover, this proposed estimator maintains the double robustness property in the original sense while enhancing robustness through a semiparametric specification. The asymptotic properties of the proposed estimators are investigated, and extensive simulation studies are conducted to evaluate its performance in finite samples and highlight its advantages over other methods. Subsequently, the proposed method is applied to the Veterans' Administrative Lung Cancer dataset and the German Breast ... |
