安德森分类统计量Andson′s classification statistic 发表评论(0) 编辑词条

Of Anderson’s classification statistic W into one of two multivariate normal populations（多元正态群体） with unknown mean vectors and unknown common covariance matrix. In practice there is the problem of the choice of a suitable cut-off point for W when upper confidence bounds are imposed on one or both types of allocation errors. By using the Studentized form of W, it is shown that a cut-off point can be easily chosen such that the probability of one type of conditional error not exceeding a specified upper bound is approximately equal to any desired value. The case where both types of conditional errors are to be bounded with high probability is considered also.