ABSTRACT
For all-pairwise multiple comparisons (MCA), the parameters of primary interest are fa — p,j for all i ^ j, the k(k — l)/2 pair wise differences of treatment means.
5.1 Balanced one-way model
Suppose under the iih treatment a random sample YH, Y^, . . . , Y{n of size n is taken, where between the treatments the random samples are independent. Then under the usual normality and equality of variances assumptions, we have the one-way model
Yia = fa + eia, i = l , . . . , f c , o = l , . . . ,n , (5.1) where fa is the effect of the iih treatment, i = 1, . . . , &, and en , . . . , e^n are i.i.d. normal with mean 0 and variance a2 unknown. We use the following notation
a2 = MSB = £ £(yio - Yi)2/(k(n - 1)] i=l a=l
for the sample means and the pooled sample variance, and let v — k(n — 1), the degrees of freedom associated with <j2.
