274 Chapter 7: Parameter Estimation
Now it can be shown that for integral valuesmandr
∫ 1
0
θm(1−θ)rdθ=
m!r!
(m+r+1)!
(7.8.1)
Hence, upon lettingx=
∑n
i= 1 xi
f(θ|x 1 ,...,xn)=
(n+1)!θx(1−θ)n−x
x!(n−x)!
(7.8.2)
Therefore,
E[θ|x 1 ,...,xn]=
(n+1)!
x!(n−x)!
∫ 1
0
θ^1 +x(1−θ)n−xdθ
=
(n+1)!
x!(n−x)!
(1+x)!(n−x)!
(n+2)!
from Equation 7.8.1
=
x+ 1
n+ 2
Thus, the Bayes estimator is given by
E[θ|X 1 ,...,Xn]=
∑n
i= 1
Xi+ 1
n+ 2
As an illustration, if 10 independent trials, each of which results in a success with probability
θ, result in 6 successes, then assuming a uniform (0, 1) prior distribution onθ, the Bayes
estimator ofθis 7/12 (as opposed, for instance, to the maximum likelihood estimator
of 6/10). ■
REMARK
The conditional distribution ofθgiven thatXi=xi,i=1,...,n, whose density function
is given by Equation 7.8.2, is called the beta distribution with parameters
∑n
i= 1 xi+1,
n−
∑n
i= 1 xi+1. ■
EXAMPLE 7.8b SupposeX 1 ,...,Xnare independent normal random variables, each having
unknown meanθand known varianceσ 02 .Ifθis itself selected from a normal population
having known meanμand known varianceσ^2 , what is the Bayes estimator ofθ?
SOLUTION In order to determineE[θ|X 1 ,...,Xn], the Bayes estimator, we need first
determine the conditional density ofθgiven the values ofX 1 ,...,Xn. Now
f(θ|x 1 ,...,xn)=
f(x 1 ,...,xn|θ)p(θ)
f(x 1 ,...,xn)