IEU Seminar: Qingyuan Zhao, Postdoctoral Fellow, Department of Statistics, University of Pennsylvania
MRC INTEGRATIVE EPIDEMIOLOGY UNIT (IEU)
Tuesday, 28th November, 2017 : 12.00 – 13.00
Room OS6, Second Floor, Oakfield House
Postdoctoral Fellow, Statistics Department,
Wharton School, University of Pennsylvania
Two-Sample Instrumental Variable Analysis: Challenges and Some Progress
(Joint work with Jingshu Wang, Jack Bowden and Dylan Small)
Instrumental variable (IV) analysis is a method of estimating causal relationships between an exposure variable and an outcome variable when there is unmeasured confounding bias. IV is widely used in econometrics and medical studies. In many emerging applications, the relevant variables are not observed in the same sample. Often, we only observe the IV and the exposure variable in sample A and the IV and the outcome variable in sample B.
In the first part of the talk, I will examine the validity of IV analysis when the two samples are from different populations. The most interesting findings are: 1. Besides the usual IV assumptions, additional untestable assumptions are generally needed; 2. The IV analysis is no longer robust to misspecified instrument-exposure model; 3. The two stage least squares is not asymptotically efficient.
In the second part of the talk, I will consider the two-sample Mendelian randomization problem in which genetic variants are used as IV. Statistically, this can be formulated as an "errors-in-variable" regression problem. I will consider a profile-likelihood estimator and its "many weak IV" asymptotics. If time allows I will discuss how to modify the estimator to be robust against pleiotropy effects (violation of exclusion restriction).
Dr Zhao is a postdoctoral fellow in the Statistics Department of the Wharton School, University of Pennsylvania (mentored by Dylan Small and Sean Hennessy). He is broadly interested in causal inference, high dimensional statistics and applied statistics. He received my Ph.D. in Statistics from Stanford University (advised by Trevor Hastie) and his B.S. in Mathematics from the Special Class for the Gifted Young (SCGY), University of Science and Technology of China (USTC) in 2011. He has also worked at eBay and Google.