Statistics Seminar（2017-01）

Topic:Identifying Causal Effects With Proxy Variables of an Unmeasured Confounder

Speaker:Wang Miao, Beijing International Center for Mathematical Research, Peking University

Time:Thursday, 5 January,14:00-15:00

Place:Room 217, Guanghua Building 2

Abstract:

Suppose we are interested in a causal effect that is confounded by an unobserved variable. Suppose however one has available proxy variables of the confounder. We show that, with at least two independent proxy variables satisfying a certain rank condition, the causal effect is nonparametrically identified, even if the error mechanism, i.e., the conditional distribution of the proxies given the confounder, may not be identified. Our result generalizes the identification strategy of Kuroki & Pearl (2014) that rests on identification of the error mechanism. When only one proxy of the confounder is available, or the required rank condition is not met, we develop a strategy to test the null hypothesis of no causal effect.

Introduction:

Wang Miao is a Ph.D. candidate in the Department of Probability and Statistics at Peking University under the supervision of Professor Zhi Geng and expect to graduate in June 2017. His primary research centers around nonignorable missing data analysis, doubly robust estimation, and causal inference with unmeasured confounding.

Your participation is warmly welcomed!