* Split plot data with CR whole plot design ; data split ; input blk a b y @@ ; cards ; 1 1 1 3 1 1 2 4 1 1 3 7 1 1 4 7 2 1 1 6 2 1 2 5 2 1 3 8 2 1 4 8 3 1 1 3 3 1 2 4 3 1 3 7 3 1 4 9 4 1 1 3 4 1 2 3 4 1 3 6 4 1 4 8 5 2 1 1 5 2 2 2 5 2 3 5 5 2 4 10 6 2 1 2 6 2 2 3 6 2 3 6 6 2 4 10 7 2 1 2 7 2 2 4 7 2 3 5 7 2 4 9 8 2 1 2 8 2 2 3 8 2 3 6 8 2 4 11 ; proc print ; run ; proc glm ; class blk a b ; model y = a blk(a) b a*b ; random blk(a) / test ; run ; /* Notice that the F tests agree between GLM and MIXED, but the standard errors of the means are different. Proc MIXED has correct standard errors */ proc mixed ; class blk a b ; model y = a b a*b / ddfm = kr ; random blk(a) ; lsmeans a b a*b / diff ; run ;