Julien Condé: Limiting distribution of maximum off-diagonal entry of sample covariance matrix from equi-correlated gaussian sample

par Julien Condé (Institut Denis Poisson, CNRS)

jeudi 29 janv. 2026, 10:30 → 12:30 Europe/Paris

Salle de séminaire (IDP-Orléans)

One is interested in the covariance structure of high-dimensional distributions. Here, we consider the equi-correlated gaussian distributions. To propose a test statistic to test alternatives between different equi-correlated cases, we derive the limiting distribution of the maximum off-diagonal entry of the sample covariance matrix, as both the sample size $n$ and dimension $p$ grow large. We improve upon the result of Fan \& Jiang, reaching the ultra-high dimensional regime where $p$ can be as large as $e^{o(n^{2/3})}$.