Distribution and correlation-free two-sample test of high-dimensional means (Q2196222)

The authors propose a two-sample test for comparing high-dimensional means that requires neither distributional nor correlational assumptions, besides some weak conditions on the moments and tail properties of the elements in the random vectors. This two-sample test called ``distribution and correlation-free (DCF) two-sample mean test'' is based on a nontrivial extension of the one-sample central limit theorem. The proposed test does not require the independently and identically distributed assumption. Weaker moments and tail conditions are posed than in the existing methods. The test allows highly unequal sample sizes. It has consistent power behavior under fairly general alternative. Simulated and real data examples demonstrate good numerical performance compared with existing methods.