A Simple Multivariate Test for One-Sided Alternatives
提出一种针对多元正态数据的简单检验方法，当均值向量倾向于正值时具有良好检验功效，并证明其第一类错误率等于α，给出了检验功效的紧界。
Abstract A simple test for multivariate normal data is proposed that has good power for alternatives where the mean vector tends to be positive. The test rejects if the quadratic form of the sample mean vector exceeds its 2α critical value and the sum of the elements of the mean vector exceeds zero. The proposed test is shown to have type I error rate equal to α whether or not the covariance matrix is known. Tight bounds on the power of the proposed test are provided, and the test is compared to three likelihood ratio tests.