Generalized Likelihood Ratio Test of Separate Families
主 题: Generalized Likelihood Ratio Test of Separate Families
报告人: Xiaoou Li (University of Minnesota)
时 间: 2017-05-19 14:00-15:00
地 点: 理科1号楼1418
Abstract: Generalized likelihood ratio test (GLRT) is a classic approach for composite hypothesis testing. In this talk, I will discuss three related problems. First, I will present a large deviation result of a GLRT for testing non-overlapping parametric families. Second, I will introduce the GLRT in sequential analysis and discuss its asymptotic properties. Third, I will present a level-triggered GLRT under a decentralized sequential design. In theory, the proposed procedures are shown to be asymptotically optimal. These procedures can be applied to various disciplines including education, engineering, and e-commerce.
About the speaker: Xiaoou Li is Assistant Professor in the School of Statistics at the University of Minnesota. She obtained her Ph.D. in Statistics from Columbia University in 2016 and B.S. in Mathematics from Peking University in 2011. Her research interests include sequential analysis and adaptive designs, large deviation and rare event analysis, latent variable modeling, and their applications in education and psychology.
