报告题目：Robust Inference for Misspecified Threshold Regression and Regression Tree Analysis

报告人：于萍 副教授 （香港大学）

报告时间：2023年5月31日上午 10:00-11:00

报告地点：学院会议室213

报告摘要:

In this paper, we develop the asymptotic theory for threshold regression under misspecification, which is especially useful in the regression tree analysis of machine learning. First, we provide a thorough characterization of the asymptotic distribution of the least square estimator, which integrates some fragmented asymptotic results of threshold regression in the literature into one unified framework of misspecification. The asymptotic distribution depends on the fitted threshold regression model being discontinuous or continuous and also on the rate of the limit objective function shrinking to zero in the direction of threshold parameter. Second, we provide a LR-based inference method for the threshold point, which can be treated as a misspecification-robust extension of the method in Hansen (2000, Econometrica, 68, 575-603). Two empirical applications illustrate the usefulness of our new inference method.

报告人简介：香港大学经济及工商管理学院副教授，博士生导师。2009年博士毕业于美国威斯康星大学麦迪逊分校，2009-2014年在新西兰奥克兰大学做助理教授。主持过香港港府基金的项目4项，已发表学术论文17篇，其中多篇发表于Journal of Econometrics, Econometric Theory和Journal of Business and Economic Statistics。

研究兴趣：微观计量经济学特别是门限回归，处理效应评估和机器学习。具体研究工作见个人主页：https://pweb.fbe.hku.hk/~pingyu/index.htm。