报告人：王景丽（南开大学）

时间：2022年05月19日10:30开始

腾讯会议ID：870 978 644 

点击链接入会: https://meeting.tencent.com/dm/VjLFjR4l3c1b 

摘要：Instrumental variable (IV) methods are widely used to identify causal effects in structural equation models with endogenous explanatory variables. Often the conditions for instruments validity are only partially plausible and some valid instruments may only offer a weak support to the causal structure. Including weak or partially invalid instruments in an analysis may lead to non-negligible biases. Allowing the proportion of invalid IV to vary in this paper, we present an approach to sensitivity analysis for the IV method which examines the sensitivity of inferences to violations of IV validity. In particular, we construct a consistent interval estimation based on the classical Fieller’s method since the causal parameter from the structural equations is a ratio of two parameters. We provide necessary theoretical and empirical supports to the proposed sensitivity analysis. A real data for home price is used to illustrate our method.

简介：王景丽，南开大学统计与数据科学学院讲师，2019年博士毕业于新加坡国立大学，主要从事生物统计及网络结构数据的研究。在Journal of Business & Economic Statistics、Statistics in Medicine等国际学术期刊上发表论文多篇，目前主持国家自然科学基金青年项目1项。

邀请人：夏小超

欢迎广大师生积极参与！