ON THE COMPARISON OF SEVERAL MEAN VALUES: AN ALTERNATIVE APPROACH

Abstract

Let yt (t = 1, 2,..., k) be statistical quantities normally and independently distributed with means fit and variances A,erf, respectively, where the A, are known constants but nothing is known about /it and cr\\. Suppose that the data provide estimates sf of the <T \\ which are distributed, respectively, as;tfcrf// ( where/ ( is the number of degrees of freedom of x2-