9.4Statistical Inferences About the Regression Parameters 365
To compute the value of the test statistic, we first compute the values ofSxx,SYY, andSxY.
A hand calculation yields that
Sxx=700, SYY=21.757, SxY=− 119
Using Equation 9.3.4 gives
SSR=[SxxSYY−SxY^2 ]/Sxx
=[700(21.757)−(119)^2 ]/700=1.527
Because
B=SxY/Sxx=−119/700=−.17
the value of the test statistic is
TS=
√
5(700)/1.527(.17)=8.139
Since, from Table A2 of the Appendix,t.005,5=4.032, it follows that the hypothesis
β=0 is rejected at the 1 percent level of significance. Thus, the claim that the mileage
does not depend on the speed at which the car is driven is rejected; there is strong evidence
that increased speeds lead to decreased mileages. ■
A confidence interval estimator forβis easily obtained from Equation 9.4.2. Indeed,
it follows from Equation 9.4.2 that for anya,0<a<1,
P
−ta/2,n− 2 <
√
(n−2)Sxx
SSR
(B−β)<ta/2,n− 2
= 1 −a
or, equivalently,
P
{
B−
√
SSR
(n−2)Sxx
ta/2,n− 2 <β<B+
√
SSR
(n−2)Sxx
ta/2,n− 2
}
= 1 −a
which yields the following.
Confidence Interval forβ
A 100(1−a) percent confidence interval estimator ofβis
(
B−
√
SSR
(n−2)Sxx
ta/2,n− 2 ,B+
√
SSR
(n−2)Sxx
ta/2,n− 2
)