B0397
Title: Novel bootstrap tests for parametric structures of high dimensional covariances
Authors: Lyudmila Sakhanenko - Michigan State University (United States) [presenting]
Nilanjan Chakraborty - Missouri University of Science and Technology (United States)
David Zhu - Michigan State University (United States)
Abstract: New bootstrap tests for parametric structures of the underlying covariance matrices based on collections of either independent high-dimensional vectors or data from linear regression in high dimensions are proposed. They are versatile and can be tuned by choosing a class of matrices. They do not require computationally costly precision matrix estimation and they work well for both sparse and dense datasets. Their performance is demonstrated theoretically, compared with existing techniques on simulated datasets, and their applicability via a real neuroimaging dataset is illustrated based on diffusion tensor imaging type of MRI.
