讲座主题：Model Averaging with High Frequency Regression

主讲嘉宾：陈大川 新加坡管理大学经济学院助理教授

讲座地点：中央财经大学沙河校区学院楼11号楼308

讲座时间：2026年11月10日(周二)13:30

嘉宾简介：2024年7月起任新加坡管理大学（SMU）经济学院经济学与统计学助理教授。他于2019年5月获得伊利诺伊大学芝加哥分校（UIC）工商管理博士学位，曾先后在芝加哥大学统计系（2015–2016年）和香港科技大学资讯、商业统计及营运学系（2023–2024年）担任访问学者。2018年，他获得芝加哥大学Stevanovich金融数学中心颁发的Stevanovich学生奖学金。他的研究成果发表于统计学与计量经济学领域的多本顶级期刊，包括Journal of the American Statistical Association, Journal of Econometrics,和Journal of Business and Economic Statistics。

内容摘要：High frequency regression models are increasingly popular in economics and related fields, but selecting the best model among a growing number of high frequency factors is challenging. This paper proposes a penalized model averaging (PMA) approach for high frequency regression, which allows for time-varying and stochastic regression coefficients and covariance structures. The proposed PMA selects data-driven weights to account for parameter uncertainty and model misspecification. We establish the asymptotic optimality of PMA when all candidate models are misspecified and demonstrate that it asymptotically assigns all weights to correctly specified models when included in the model set. These properties hold for both fixed and diverging numbers of candidate models, with the latter providing insights into high-dimensional coefficient estimation. Simulations and empirical studies confirm the validity and superior performance of the proposed methodology.