270 Chapter 7: Parameter Estimation
EXAMPLE 7.7c LetX 1 ,...,Xndenote a sample from a uniform (0,θ) distribution, where
θis assumed unknown. Since
E[Xi]=
θ
2
a “natural” estimator to consider is the unbiased estimator
d 1 =d 1 (X)=
2
∑n
i= 1
Xi
n
SinceE[d 1 ]=θ, it follows that
r(d 1 ,θ)=Var(d 1 )
=
4
n
Var(Xi)
=
4
n
θ^2
12
since Var(Xi)=
θ^2
12
=
θ^2
3 n
A second possible estimator ofθis the maximum likelihood estimator, which, as shown
in Example 7.2d, is given by
d 2 =d 2 (X)=max
i
Xi
To compute the mean square error ofd 2 as an estimator ofθ, we need to first compute
its mean (so as to determine its bias) and variance. To do so, note that the distribution
function ofd 2 is as follows:
F 2 (x)≡P{d 2 (X)≤x}
=P{max
i
Xi≤x}
=P{Xi≤x for alli=1,...,n}
=
∏n
i= 1
P{Xi≤x} by independence
=
(x
θ
)n
x≤θ
Hence, upon differentiating, we obtain that the density function ofd 2 ,is
f 2 (x)=
nxn−^1
θn
,x≤θ