*7.6Confidence Interval of the Mean of the Exponential Distribution 265
*7.6Confidence Interval of the Mean of the Exponential Distribution
IfX 1 ,X 2 ,...,Xnare independent exponential random variables each having meanθ, then
it can be shown that the maximum likelihood estimator ofθis the sample mean
∑n
i= 1 Xi/n.
To obtain a confidence interval estimator ofθ, recall from Section 5.7 that
∑n
i= 1 Xihas
a gamma distribution with parametersn,1/θ. This in turn implies (from the relationship
between the gamma and chi-square distribution shown in Section 5.8.1.1) that
2
θ
∑n
i= 1
Xi∼χ 22 n
Hence, for anyα∈(0, 1)
P
{
χ 12 −α/2,2n<
2
θ
∑n
i= 1
Xi<χα^2 /2,2n
}
= 1 −α
or, equivalently,
P
2
∑n
i= 1
Xi
χα^2 /2,2n
<θ <
2
∑n
i= 1
Xi
χ 12 −α/2,2n
= 1 −α
Hence, a 100(1−α) percent confidence interval forθis
θ∈
2
∑n
i= 1
Xi
χα^2 /2,2n
,
2
∑n
i= 1
Xi
χ 12 −α/2,2n
EXAMPLE 7.6a The successive items produced by a certain manufacturer are assumed to
have useful lives that (in hours) are independent with a common density function
f(x)=
1
θ
e−x/θ,0<x<∞
If the sum of the lives of the first 10 items is equal to 1,740, what is a 95 percent confidence
interval for the population meanθ?
* Optional section.