294 Chapter 8:Hypothesis Testing
SinceX=
∑n
i= 1 Xi/nis a natural point estimator ofμ, it seems reasonable to accept
H 0 ifXis not too far fromμ 0. That is, the critical region of the test would be of the form
C={X 1 ,...,Xn:|X−μ 0 |>c} (8.3.1)
for some suitably chosen valuec.
If we desire that the test has significance levelα, then we must determine the critical
valuecin Equation 8.3.1 that will make the type I error equal toα. That is,cmust be
such that
Pμ 0 {|X−μ 0 |>c}=α (8.3.2)
where we writePμ 0 to mean that the preceding probability is to be computed under the
assumption thatμ=μ 0. However, whenμ=μ 0 ,Xwill be normally distributed with
meanμ 0 and varianceσ^2 /nand soZ, defined by
Z≡
X−μ 0
σ/
√
n
will have a standard normal distribution. Now Equation 8.3.2 is equivalent to
P
{
|Z|>
c
√
n
σ
}
=α
or, equivalently,
2 P
{
Z>
c
√
n
σ
}
=α
whereZis a standard normal random variable. However, we know that
P{Z>zα/2}=α/2
and so
c
√
n
σ
=zα/2
or
c=
zα/2σ
√
n
Thus, the significance levelαtest is to rejectH 0 if|X−μ 0 |>zα/2σ/
√
nand accept
otherwise; or, equivalently, to
reject H 0 if
√
n
σ
|X−μ 0 |>zα/2
accept H 0 if
√
n
σ
|X−μ 0 |≤zα/2
(8.3.3)