Variation Independent Parameterizations of Binary Causal Models
主 题: Variation Independent Parameterizations of Binary Causal Models
报告人: Linbo Wang (Harvard University)
时 间: 2017-07-05 14:00-15:00
地 点: 理科一号楼1114
Abstract: A common problem in formulating models for causal effects under binary causal models is the variation dependence between the causal parameters and the commonly-used nuisance models. Such causal models include the structural mean model for the conditional relative risk and risk difference, the binary instrumental variable model, the binary structural nested mean model, and so forth. We address this problem by proposing novel nuisance models based on the conditional odds product. These novel nuisance models facilitate maximum-likelihood estimation, but also permits doubly-robust estimation for the causal parameters of interest.
About the speaker: Linbo Wang is a postdoctoral fellow in the Department of Biostatistics at Harvard University, mentored by Professors Eric Tchetgen Tchetgen and James Robins. He received his Ph.D. degree in Biostatistics from the University of Washington, advised by Professors Thomas Richardson and Xiao-Hua Andrew Zhou. His main research interests include causal inference, graphical models, missing data, and robust inference in infinite-dimensional models.