Robust two-sample test of high-dimensional mean vectors under dependence (Q1755128)

Consider testing the equality of the means \(\mu_1,\mu_2\) of two normal distributed random vectors \(X,Y\) having the same covariance matrix \(\Sigma\). Here, the problem is treated that the dimension \(p\), of the random vectors is comparable to, or even larger than the size \(n\) of the samples of \(X,Y\). Since, in this case, standard tests, as Hotelling's \(T^2\) test, have a low performance and are sensitive to outliers, a more robust test statistic is suggested based on the Kolmogorov distance, adopting trimmed means and robust precision matrix estimators. A simulation study is given.