Estimation of deviation for random covariance matrices (Q2335874)

The authors consider random covariance matrices of the form \(W=M^* M\), where \(M\) is a \(p\times n\)-random matrix whose entries are independent (not necessarily identically distributed) random variables with zero mean, unit variance, and uniformly bounded fourth moments. They prove an explicit estimate on the probability that certain distance between the empirical spectral distribution of \(W/n\) and the Marchenko-Pastur law exceeds some given number.