Jun Wang

We propose a semiparametric method to estimate average treatment effects in observational studies based on the assumption of unconfoundedness. Assume that the propensity score model and outcome model are a general single index model, which are estimated by the kernel method and the unknown index parameter is estimated via linearized maximum rank correlation method. The proposed estimator is computationally tractable, allows for large dimension covariates and not involves the approximation of link functions. We showed that the proposed estimator is consistent and asymptotically normally distributed. In general, the proposed estimator is superior to existing methods when the model is incorrectly specified. We also provide an empirical analysis on the average treatment effect and average treatment effect on the treated of 401(k) eligibility on net financial assets.

Treatment effects estimation is one of the crucial mainstays in medical and epidemiological studies. Ignorance of the existence of confounders may result in biased estimators. The issue will become more serious and complicated if the treatment is endogenous (i.e., the presence of unobserved confounders). In this article, we propose a new treatment effects estimator for binary treatments in observational studies in the presence of unobserved confounders. The proposed estimator is consistent and asymptotically normally distributed. A statistic is also developed for testing the existence of treatment effects. Simulation studies show that the proposed estimator is stable for various unobserved confounding settings and the distribution of error terms. Finally, we apply our proposed methodologies to a low birthweight data set which yields different conclusions with and without the consideration of possible unobserved confounders.