Suppose we have two independent high-dimensional samples:
$$
\boldsymbol{y}_{i1},\ldots,\boldsymbol{y}_{in_i}, \;\operatorname{are \; i.i.d. \; with}\; \operatorname{E}(\boldsymbol{y}_{i1})=\boldsymbol{\mu}_i,\; \operatorname{Cov}(\boldsymbol{y}_{i1})=\boldsymbol{\Sigma},i=1,2.
$$
The primary object is to test
$$H_{0}: \boldsymbol{\mu}_1 = \boldsymbol{\mu}_2\; \operatorname{versus}\; H_{1}: \boldsymbol{\mu}_1 \neq \boldsymbol{\mu}_2.$$
Srivastava and Du (2008) proposed the following test statistic:
$$T_{SD} = \frac{n^{-1}n_1n_2(\bar{\boldsymbol{y}}_1 - \bar{\boldsymbol{y}}_2)^\top \boldsymbol{D}_S^{-1}(\bar{\boldsymbol{y}}_1 - \bar{\boldsymbol{y}}_2) - \frac{(n-2)p}{n-4}}{\sqrt{2 \left[\operatorname{tr}(\boldsymbol{R}^2) - \frac{p^2}{n-2}\right] c_{p, n}}},$$
where \(\bar{\boldsymbol{y}}_{i},i=1,2\) are the sample mean vectors, \(\boldsymbol{D}_S\) is the diagonal matrix of sample variance, \(\boldsymbol{R}\) is the sample correlation matrix and \(c_{p, n}\) is the adjustment coefficient proposed by Srivastava and Du (2008).
They showed that under the null hypothesis, \(T_{SD}\) is asymptotically normally distributed.