Title: Gaussianity test for high-dimensional data
Authors: Hao Chen - University of California at Davis (United States) [presenting]
Yin Xia - Fudan University (China)
Abstract: Many high-dimensional data analysis tools require the data have Gaussian or sub-Gaussian tails. We here provide a general framework for testing whether the data have Gaussian tail or heavier tail. The method extends from graph-based two-sample tests and work when a reasonable covariance matrix estimation is possible. Under some mild conditions on the covariance matrix estimation, the test is consistent against all distributions with tail heavier than the Gaussian distribution.