Exercise 8.1 Using a randomized complete block design is supposed to reduce the variability of treatment comparisons. If the randomized complete block model is taken as yi j = μ +αi+β j +ei j, ei js i.i.d. N(0,σ 2), i = 1, . . . ,a, j = 1, . . . ,b, argue that the corresponding variance for a completely randomized design should be σ 2+Σbj

=1(μ j −μ¯·)2/b, where μi = μ +β j.