306 Chapter 8:Hypothesis Testing
Now whenσ^2 is no longer known, it seems reasonable to estimate it by
S^2 =
∑n
i= 1
(Xi−X)^2
n− 1
and then to rejectH 0 when
∣
∣∣
∣∣
X−μ 0
S/
√
n
∣
∣∣
∣∣
is large.
To determine how large a value of the statistic
∣
∣∣
∣∣
√
n(X−μ 0 )
S
∣
∣∣
∣∣
to require for rejection, in order that the resulting test have significance levelα, we must
determine the probability distribution of this statistic whenH 0 is true. However, as shown
in Section 6.5, the statisticT, defined by
T=
√
n(X−μ 0 )
S
has, whenμ=μ 0 ,at-distribution withn−1 degrees of freedom. Hence,
Pμ 0
{
−tα/2,n− 1 ≤
√
n(X−μ 0 )
S
≤tα/2,n− 1
}
= 1 −α (8.3.11)
wheretα/2,n− 1 is the 100α/2 upper percentile value of thet-distribution withn−1 degrees
of freedom. (That is,P{Tn− 1 ≥tα/2,n− 1 }=P{Tn− 1 ≤−tα/2,n− 1 }=α/2 whenTn− 1
has at-distribution withn−1 degrees of freedom.) From Equation 8.3.11 we see that the
appropriate significance levelαtest of
H 0 :μ=μ 0 versus H 1 :μ=μ 0
is, whenσ^2 is unknown, to
accept H 0 if
∣∣
∣
∣∣
√
n(X−μ 0 )
S
∣∣
∣
∣∣≤tα/2,n− 1
reject H 0 if
∣∣
∣∣
∣
√
n(X−μ 0 )
S
∣∣
∣∣
∣
>tα/2,n− 1
(8.3.12)