Speaker：Yong He (Shandong University)

Time：2024-5-31, 10:30

Location：Conference Room 105 at Experiment Building at Haiyun Campus

Abstract:

In this talk, I will introduce a new type of Kendall's tau for robust statistics, names as matrix-type Kendall's tau, which generalize the spatial Kendall's tau (Marden, 1999) in the literature to deal with random matrix elliptical observations. I will elaborate on its use in robust estimation for both factor model and principal eigenvectors (under both sparse and non-sparse settings) in High-dimensions. I will also introduce how to extend the tool to deal with high-order tensors. We also develop an R package “MKendall” which is available at CRAN.