国产探花 学术报告[2025] 093号

(高水平大学建设系列报告1115号)

报告题目：A data-driven approach for independence testing in multivariate time series based on Chatterjee's rank correlation

报告人：王国长 教授（暨南大学）

报告时间：2025年10月14日下午16:00-17:00

报告地点：校友广场303会议室 　　　　　　　　　　　　　　　　　　　　　　　　　

报告摘要：Verifying whether a multivariate time series follows an independent and identically distributed (i.i.d.) sequence is a critical issue.This paper introduces a novel method for testing this assumption by analyzing the dependency structure inherent in the data. Specifically, we utilize Chatterjee's rank correlation to develop a new analytical tool, termed the auto-Chatterjee's rank correlation matrix (ACRCM). Each element of the ACRCM quantifies Chatterjee's rank correlation, effectively capturing nonlinear dependencies within the observed series. We develop an ACRCM-based statistic with fixed order for testing independence in multivariate time series, which asymptotically follows a chi-squared distribution under the assumption of independence. Furthermore, we introduction a data-driven approach for automatically determining the optimal order based on the data characteristics.This data-driven approach offers three key advantages:first, it eliminates the need for manually specifying the order, as the optimal order is automatically selected based on the data;second, under the null hypothesis, the selected order is one, and the chi-square distribution has degrees of freedom correspond to the square of the data dimension;third, the proposed data-driven approach demonstrates superior sensitivity to detecting high-order dependencies.We rigorously derive the asymptotic properties of the proposed method and validate its effectiveness through extensive simulation experiments.

报告人简介：王国长，暨南大学经济学院统计与数据科学系，教授、博士生导师。主要研究方向为函数型数据，时间序列和机器学习，至今在JoE、JBES、Sinica和Scandinavian Journal of Statistics等重要学术期刊发表论文30余篇。主持国家级项目4项，省部级项目4项。任中国旅游大数据协会，副理事；广东省现场统计协会常务理事，秘书长。

欢迎师生参加！

邀请人：王江洲

　　　　　　　　　　　　　　　　　　　　　　　国产探花 　　　　　　　　　　　　　　　　　　　　　 2025年9月30日