An improved asymptotic distribution of largest absolute off-diagonal entries of high-dimensional sample covariance matrix
时间：2019-06-03
阅读量：199次
- 演讲人： 郑术蓉教授（东北师范大学数学与统计学院）
- 时间：2019年06月13日 星期四上午9:00-
- 地点：紫金港校区管理学院行政楼14楼1417报告厅
摘要：Many papers have studied the asymptotic distribution of the statistic of largest absolute off-diagonal entries of high-dimensional sample covariance matrix when the population covariance matrix is an identity matrix. But, due to the reason of the slow convergence speed, these asymptotic distributions are still a little away from the empirical distributions of the statistic. The target of this paper is to find an improved asymptotic distribution of the statistic. The improved asymptotic distribution will be very close to the empirical distributions of the statistic although the sample size $n$ and the dimension $p$ is not very large with $p/n\rightarrow\rho\in(0, \infty)$. When the dimension and sample size are large with $p/n\rightarrow\rho\in(0, \infty)$, the improved asymptotic distribution will be almost same as the empirical distributions of the statistic.
欢迎参加！
联系人: 张立新教授 suzhonggen@zju.edu.cn
浙江大学数据科学研究中心、浙江大学数学科学学院统计学研究所
报告人简介：
郑术蓉， 东北师范大学数学与统计学院教授、博士生导师，主要从事大维随机矩阵理论及高维统计推断的研究， 如建立大维Fisher随机矩阵线性谱统计量的中心极限定理、建立大维样本相关矩阵线性谱统计量的中心极限定理、检验大维协方差矩阵的结构和相等性、检验大维相关矩阵的相等性并用于脑成像图的研究中、研究大维因子分析模型等。