Let R (β_0 β_1 β_2 . . . β_p) = SS(Model) denote the sum of squares due
to the model containing the parameters listed in parentheses. The sum
of squares for the hypothesis that a subset of β_j is zero can be obtained
by subtraction of SS(Model) for the reduced model from that for the full
model. Assume the subset of β_j being tested against zero consists of the
last k β_j . Then