The usual model for a randomized complete block design was given in Section 8.2 as Yij = /-L + Gi + (Jj + eij, i = 1, ... ,

a, j = 1, ... ,

b, Var(eij) = a 2, and Cov(eij, eiljl) = ° for (i,j) ¥=

(i',j'). The (JjS are considered as fixed block effects. Consider now a model Yij = /-L + Gi + (Jj + 'f/j + eij'

The 'f/jS are independent N(O, a~) and the eijS are independent N(O, af).

The 'f/jS and eijS are also independent. The block effects are now ((Jj + 'f/j).

There is a fixed component and a random component with mean zero in each block effect. Use the results of this section to derive an analysis for this model. Give an ANOVA table and discuss interval estimates for contrasts in the GiS.