A Test of Homogeneity for Ordered Variances
有序方差的同质性检验
SUMMARY Suppose that an estimate of variance is available from each of a set of normal populations. The situation in which there is a definite trend in the magnitude of the population variances is considered, and this is represented by assuming the ith variance to be inversely proportional to λ + iμ. A uniformly most powerful unbiased test is developed for the hypothesis μ = 0, and methods for estimating μ and λ are considered.