报告题目：A Bootstrap Method for Testing High-dimensional White Noise

主 讲 人：陈敏 中国科学院数学与系统科学研究院

报告时间：2021年11月20日（周六）10:00-11:00

报告地点：知新楼B1238

报告摘要：

For high-dimensional time series, this article proposes a new test to detect white noise that is not necessary to be assumed as independent and identically distributed. The dimension p is allowed to go to infinity when n → ∞. The test is constructed based on the extreme value of auto-correlations and cross-correlations. The distribution of the test statistic is approximated by uniform multiplier (or wild) bootstrap method that is extended in this article to accommodate high dimensional weakly stationary time series. The provided asymptotic properties of the new bootstrap method illustrate the validity of the approximation and thus the new test achieves the desirable size and power asymptotically. This accuracy of approximation is further demonstrated by the simulation study which compares the performance of the new test with that of recently developed tests designed for high dimensional time series, the conventional portmanteau test and several variations of it. The simulation results indicate that the new test outperforms in both empirical sizes and powers for finite samples with small or large p relative to n. In addition, the performance is consistent over various models with different distributions.

主讲人简介：

陈敏，中国科学院数学与系统科学研究院研究员，博士生导师，享受国务院政府特殊津贴。现任全国统计方法应用技术标准化委员会主任委员,《数理统计与管理》主编，《应用数学学报（中文版）》副主编，全国工业统计学教学研究会会长、中国现场统计研究会经济与金融统计分会理事长。曾任中国数学学会副理事长、中国统计教育学会副会长、北京大数据协会副会长。曾任中国科学院数学与系统科学研究院副院长。主要研究方向：金融统计理论与方法、非线性时间序列的统计分析，非参数统计估计和检验的大样本理论，生物统计的理论与方法，应用统计(工业统计、统计标准化、财税信息技术)，大数据分析与处理的统计理论与算法研究。出版和翻译教材和专著７部；在国内外核心学术期刊发表统计理论与应用、经济、金融和管理学论文150余篇，其中SCI和EI论文80余篇。

邀请人：陈增敬教授

主办单位：山东大学金融研究院、数学学院

欢迎各位老师同学积极参加！