主题：Testing for the symmetry of innovation in multivariate time series models
主讲人：朱柯 香港大学
主持人：王国长 XPJ官方网站
时间：2025年5月13日（周二）下午16:30-17:30
地点：XPJ官方网站石牌校区XPJ官方网站大楼（中惠楼）102室
摘要
Determining whether the innovation is symmetrically distributed or not is important for theoretical studies and empirical applications of multivariate time series models. The traditional testing methods for univariate models depend on either smoothing or martingale transformation treatment, so their extension to multivariate models is not a straightforward or desirable task. In this paper, we propose a new consistent test to examine the symmetry of innovation in multivariate time series models by using a characteristic measure. Regardless of the dimension of the model, the proposed test is easy-to-implement without involving any smoothing treatment. Under regular conditions, we establish the limiting null distribution of the test, and show the consistency and non-trivial local power of the test for a large class of alternative hypotheses. Moreover, we adopt a valid fixed-design wild bootstrap method to approximate the critical values of the test. Interestingly, we find that model estimation has a negligible effect on the limiting null distribution for certain pure conditional mean models and any pure conditional variance model, for which the critical values can be approximated by a fast-computing wild bootstrap method. Finally, we demonstrate the usefulness of the test via simulation results and real data analysis.
主讲人简介
朱柯博士2011年获得香港科技大学统计学博士学位，同年进入中国科学院数学与系统科学研究院从事研究工作，历任助理研究员、副研究员。2016年加入香港大学任助理教授、副教授。朱柯博士的研究兴趣包括时间序列分析、计量经济、金融大数据、因果推断等领域。他的一系列论文发表在Annals of Statistics, Journal of the Royal Statistical Society (Series B), Journal of American Statistics Association, Journal of Econometrics 和 Journal of Business & Economic Statistics等。
欢迎感兴趣的师生参加！
校对| 王国长
责编| 彭毅
初审| 姜云卢
终审发布| 何凌云
（来源：XPJ官方网站微信公众号）
