Test for high dimensional covariance matrices (Q1996783)

Given samples from \(p\)-dimensional i.i.d. random vectors with mean \(\mu\) and covariance matrix \(\Sigma\), statistical methods are considered for the analysis of the structure of \(\Sigma\) and its inverse (``precision matrix'') in the case of high dimensions. Especially, procedures for testing off-diagonal covariance structure and the asymptotic properties of the test statistic as well as the theoretical properties of the half-sampling estimator are studied. Properties of tests for parametric covariance matrices are presented. Moreover, for precision matrices occurring in dimensional inference problems, the detection of an existing structure or substructure is considered. Here, a testing procedure for a given substructure of the precision matrix is provided, and some theoretical properties under sub-Gaussian tail and the linear process model are discussed. Finally, simulation studies for evaluating the performance of the tests are shown.