In this article, we propose a novel test based on summation of square of the maximum and minimum column component t-tests for testing two mean vectors equality of high dimensional data in two-sample cases with sparse nonzero differences.The proposed test employs the information about the different directions for the maximum minimum column component t-tests and is prospected to gain more power than existed tests. Theoretical properties of the proposed test are investigated. Asymptotic distributions of the proposed test are derived under some regular conditions.

Extensive simulation and a real data analysis about age-dependent regulation of gene expression are conducted to confirm the merits of the novel proposed test.