In this study, we assume that the reader is familiar with the technical precision medicine terms such as "treatment" and "treatment-free" models, confounding, adjusting for a confounder, model specification, etc.
In this causal inference project, we conduct a simulation study to explore the doubly-robustness of the dWOLS estimators. In the second part, we investigate the impact of a violation of the No Unmeasured Confounding (NUC) assumption on these estimators.
To investigate whether the dWOLS estimators are doubly-robust and how the violation of the NUC assumption affects the estimators.
To explore the doubly-robustness of the dWOLS estimators.
Obtaining the optimal terms in model specifications is not realistic thus we remove the terms from model specifications and assess how the estimators behave. The sample size could result in the bias-ness, so we chose two big different sample sizes and investigated the pattern of the results.
I like to conduct a sensitivity analysis to assess how the bias in estimators can vary in a range of sensitivity parameters.