Understanding Causal Effects in Observational Studies with Instrumental Propensity Score

• Activity Number: 10
• Type: Invited
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Committee on Women in Statistics
• Author(s): Jing Cheng*+ and Winston Lin
• Companies: University of California, San Francisco and University of California, Berkeley
• Keywords: Distributional causal effect ; instrumental variable ; instrumental propensity score ; unmeasured confounding ; observational studies

Abstract

In observational studies, investigators are often concerned about measured and unmeasured confounding in the evaluation of the effect of a program/exposure/treatment. The instrumental variable (IV) approach is valuable to address unmeasured confounding and the two-stage least square (2SLS) estimate based on the two-stage linear models has been widely used in many applications. However, challenges investigators often have with the IV approach include difficulty in finding a good IV and methods for general outcomes. In this talk, we will discuss the properties and applications of instrumental propensity score to address those problems for evaluating average and distributional causal effects for general outcomes in observational studies.