San Francisco Chapter of the American Statistical Association Seminar



Prof Jing Cheng from UCSF.



Understanding Distributional and Heterogeneous Causal Effects in Observational Studies with the Instrumental Propensity Score



Jan 27th from 4:30-6pm (4:30-5pm social and 5-6pm presentation)



M112 class room at Stanford Medical school  It is in the medical center, Alway building, ground floor, next to the Dean's courtyard.


Map:  Parking on campus after 4pm is free




Abstract:  Confounding has been a main concern in evaluating treatment effects in observational studies when randomization of the treatment assignment is not feasible. To address issues with measured and unmeasured confounding in observational studies, this paper develops a unified approach to using an instrumental variable (IV) in more flexible ways to evaluate treatment effects. The approach is based on an instrumental propensity score conditional on baseline variables, which can then be incorporated in matching, regression, subclassification or weighting along with various parametric, semiparametric or nonparametric methods for the assessment of treatment effects when the standard two-stage linear models do not work well. Several properties of the instrumental propensity score will be discussed in the talk. The approach will then be illustrated using subclassification along with a semiparametric density ratio model (Anderson 1979) and empirical likelihood (Owen, 1988; Qin and Lawless, 1994). This method allows us to evaluate distributional and heterogeneous treatment effects in addition to the overall average treatment effect. Simulation studies show that the method works well. We apply our method to a study of the effects of attending a Catholic school versus a public school on subsequent wages using data from the Wisconsin Longitudinal Study (WLS; Hauser and Sewell, 2005).  Some interesting findings on distributional and heterogeneous effects will be discussed in the talk.