ABSTRACT
In this chapter, the case is considered where the treatments assigned to the experimental units may affect the variance of the responses as well as the mean. Start with the one-way means model, yij = mi + eij, for i = 1, 2, º , t, j = 1, 2, º , ni. In Chapter 1 it was assumed that the experimental errors all had the same variance; that is, the treatments were expected to possibly change the mean of the population being sampled, but not the variance. In this chapter, some methods are described for analyzing data when the treatments affect the variances as well as the mean. The types of questions that the experimenter should want to answer about the means in this setting are similar to those in Chapter 1. That is,
1) Are all means equal? 2) Can pairwise comparisons among the means be made? 3) Can a test of the hypothesis of the form Âi=1
t ci mi = a be tested and can con dence
intervals be constructed about Âi=1 t ci mi?
