Asymptotic joint distribution of extreme eigenvalues and trace of large sample covariance matrix in a generalized spiked population model (Q1996762)

As the title of this article suggests, this text brings asymptotic results about the joint distribution between the covariance matrix trace and the order statistics (specifically the extreme values) of eigenvalues of the covariance matrix in a generalized spiked population model. This topic is of practical and theoretical importance. The theoretical importance stems from the fact that no one has investigated the asymptotic joint distribution of the trace and extreme values of eigenvalues and the practical importance stems from application of this result in hypothesis tests, which in turn can be applied to real data. A central result presented in this scientific article is that the asymptotic joint distribution between the trace of the covariance matrix and extreme values of eigenvalues of the covariance matrix is asymptotically normal. Furthermore, it is proved that the asymptotic distribution of covariance matrix trace is asymptotically independent of the asymptotic distribution of extreme values of eigenvalues of the covariance matrix. This paper is accompanied by supplementary material that contains additional simulations and the more technical parts of the demonstrations of some theorems.