王会霞学术报告 题目：Quantile Regression with Unknown Parameter Heterogeneity 报告人：王会霞 教授（美国乔治华盛顿大学） 地点：宁静楼110室 时间：2017年12月22日（星期三）13:30-14:30 报告摘要 In this talk, I will discuss quantile regression with group-specific parameters when the group membership is unknown. Identifying groups with homogeneous parameters can be viewed as a model-based clustering problem. I will first introduce a concave fusion penalization method that can perform simultaneous parameter estimation and group identification. The method relies on penalizing the differences of parameters between pairs of units and thus could be computationally costly for large samples. Next I will introduce an iterative two-step algorithm using a similar idea of k-means clustering to identify groups and estimate group-specific parameters for panel data. In practice the signal differentiating groups may vary across quantiles though the group membership may be common. It remains unclear which quantile is preferable or should one combine information across quantiles to perform clustering. I will present a stability measure and show how it helps to choose among single quantiles and the composite quantile. Finally, I will discuss the asymptotic properties of the proposed methods, and present some numerical results. This is a joint work with Yingying Zhang and Zhongyi Zhu from Fudan University 个人简介 王会霞教授，复旦大学本科、硕士，美国伊利诺伊大学香槟分校博士，师从著名统计学家何旭铭教授。博士毕业后曾任教于北卡州立大学统计系，现为乔治华盛顿大学统计系终身正教授、复旦大学统计系特聘东方学者。王会霞教授的研究有分位数回归、极值理论、生物信息学、非参数（半参数）回归、统计推断、变量选择、生存分析、纵向数据、空间数据分析、测量误差模型、缺失数据分析等。目前已在统计学四大顶级期刊JRSSB，Annals of Statistics，JASA，Biometrika发表论文十余篇，SCI期刊论文近60篇，是国际统计界的青年领军人物。 欢迎各位参加！