Exercise 5.8.4 Suppose that in a balanced one-way ANOVA the group means

¯y1·, . . . , ¯yt· are not independent but have some nondiagonal covariance matrix V.

How can Tukey’s HSD method be modified to accommodate this situation?

Exercise 5.8.5 For an unbalanced one-wayANOVA, give the contrast coefficients for the contrast whose sum of squares equals the sum of squares for groups. Show the equality of the sums of squares. Hint: Recall that in Exercise 4.2 we found the form of MαY and that equation (4.2.2) allows one to read off the coefficients of any contrast determined by a vector in C(Mα).