(d) Fit the two models and ¯nd the coe±cients, and residual variance in each case. For the ¯rst model, use (5.15) to estimate the residual variance. But for the second model, because there are four parameters, the residual degrees of freedom is n ¡ 4. So the residual variance ¾2

" can be estimated using the following equation (similar to (5.15))

s2e

=

1 n ¡ 4 Xn i=1 e2i

: