Scribd
Upload
3K views

An Introduction To Signal Detection and Estimation - Second Edition Chapter IV: Selected Solutions

An Introduction to Signal Detection and Estimation - Second Edition Chapter IV: Selected Solutions

Uploaded by

Shakil Ahmed Chowdhury
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
3K views

An Introduction To Signal Detection and Estimation - Second Edition Chapter IV: Selected Solutions

An Introduction to Signal Detection and Estimation - Second Edition Chapter IV: Selected Solutions

Uploaded by

Shakil Ahmed Chowdhury
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

An Introduction to Signal Detection and

Estimation - Second Edition


Chapter IV: Selected Solutions
H. V. Poor
Princeton University
April 26, 2005

Exercise 1:
a.

M AP (y) = arg max log |y| log = 1.


1e

b.

e

e|y| d
1
e|y| e1e|y|
M M SE (y) = 1 e |y|
=
+ |y|
d
|y|
e
ee|y|
1 e

Exercise 3:
y e e
y e(+1) (1 + )y+1
w(|y) =  y =
.
d
y!
0 e e
So:

1  y+1 (+1)
y+1

M M SE (y) =
e
d(1 + )y+1 =
;
y! 0
+1


y
;
M AP (y) = arg max y log ( + 1) =
>0
+1
and ABS (y) solves
 ABS (y)
1
w(|y)d =
2
0
which reduces to

k
y
ABS (y)

1
= eABS (y) .
k!
2
k=0
Note that the series on the left-hand side is the truncated power series expansion of
exp{ABS (y)} so that the ABS (y) is the value at which this truncated series equals half
of its untruncated value when the truncation point is y.
1

Exercise 7:
We have

p (y) =
and

w() =
Thus,

ey+ if y
;
0
if y <
1 if (0, 1)
.
0 if  (0, 1)

e
ey+
= min{1,y}
,
w(|y) =  min{1,y}
e
1
ey+ d
0

for 0 min{1, y}, and w(|y) = 0 otherwise. This implies:


 min{1,y}

M M SE (y) =

e d
[min{1, y} 1]emin{1,y} + 1
=
;
emin{1,y} 1
emin{1,y} 1

M AP (y) = arg
and

 ABS (y)

max

0min{1,y}

e d =

which has the solution

= min{1, y};


1  min{1,y}
e
1
2

emin{1,y} + 1

ABS (y) = log


.
2

Exercise 8:
a. We have
w(|y) =

ey+ e
1
y
= , 0 y,
y
e
y
0 d

and w(|y) = 0 otherwise. That is, given y, is uniformly distributed on the interval
[0, y]. From this we have immediately that M M SE (y) = ABS (y) = y2 .
b. We have
M M SE = E {V ar(|Y )} .
Since w(|y) is uniform on [0, y], V ar(|Y ) =
p(y) = ey

 y

Y2
.
12

We have

d = yey , y > 0,

from which
M M SE =

3!
1  3 y
1
E{Y 2 }
y e |dy =
=
= .
12
12 0
12
2
2

c. In this case,
p (y) =

eyk + , if 0 < < min{y1 , . . . , yn }.

k=1

from which

M AP (y) = arg

max

0<<min{y1 ,...,yn }

exp

 n


yk + (n 1)

= min{y1 , . . . , yn }.

k=1

Exercise 13:
a. We have
p (y) = T (y) (1 )(nT (y)) ,
where
T (y) =

n


yk .

k=1

Rewriting this as
p (y) = C()eT (y)
with = log(/(1 )), and C() = en we see from the Completeness Theorem for
Exponential Families that T (y) is a complete sucient statistic for and hence for .
(Assuming ranges throughout (0, 1).) Thus, any unbiased function of T is an MVUE for
. Since E {T (Y )} = n, such an estimate is given by
n
T (y)
1
yk .
=
M V (y) =
n
n k=1

b.

M L (y) = arg max T (y) (1 )(nT (y)) = T (y)/n = M V (y).


0<<1

Since the MLE equals the MVUE, we have immediately that E {M L (Y )} = . The
variance of M L (Y ) is easily computed to be (1 )/n.
c. We have
2
T (Y ) n T (Y )
log p (Y ) = 2
,
2

(1 )2
from which
I =
The CRLB is thus
CRLB =

n
n
n
+
=
.
1
(1 )

1
(1 )
=
= V ar (M L (Y )).
I
n

Exercise 15:
We have

e y
,
y!

p (y) =
Thus

y 0, 1, . . ..

y
log p (y) =
( + y log ) = 1 + ,

and

2
y
log
p
(y)
=

< 0.

2
2

So

M L (y) = y.
Since Y is Poisson, we have E {M L (Y )} = V ar (M L (Y )) = . So, M L is unbiased.
Fishers information is given by


I = E

2
E {Y }
1
log p (Y ) =
= .
2
2

So the CRLB is , which equals V ar (M L (Y )). (Hence, the MLE is an MVUE in this
case.)

Exercise 20:
a. Note that Y1 , Y2 , . . . , Yn , are independent with Yk having the N (0, 1 + s2k ) distribution.
Thus,


n


1
yk2

2
log p (y) =
log(1 + sk )

2
2(1 + s2k )
k=1


n
1
yk2 s2k
s2k
=

,
2 k=1 1 + s2k (1 + s2k )2

from which the likelihood equation becomes


1 M L (y)s2k )
= 0.
(1 + M L (y)s2 )2

n

s2k (yk2
k=1

b.

I = E

n

s4k
2
s4k E {Yk2 }
log p (Y ) =

2 3
2
2(1 + k2 )2
k=1 (1 + sk )

n
1
s4k
.
2 k=1 (1 + s2k )2

So the CRLB is

n

s4k
k=1 (1+s2 )2
k

c. With s2k = 1, the likelihood equation yields the solution




M L (y) =

n
1
y 2 1,
n k=1 k

which is seen to yield a maximum of the likelihood function.


d. We have


n


 

1
2
E M L (Y ) =
E Yk
1 = .
n k=1
Similarly, since the Yk s are independent,
n
n


 
1 
1 
2(1 + )2
.
V ar Yk2 = 2
2(! + )2 =
V ar M L (Y = 2
n k=1
n k=1
n

Thus, the bias of the MLE is 0 and the variance of the MLE equals the CRLB. (Hence,
the MLE is an MVUE in this case.)

Exercise 22:
a. Note that
p (y) = exp

 n


log F (yk )/ ,

k=1

which impies that the statistic nk=1 log F (yk ) is a complete sucient statistic for via
the Completeness Theorem for Exponential Families. We have
E

 n


n
log F (Yk ) = nE {log F (Y1 )} =
log F (Y1 ) [F (y1 )](1)/ f (y1 )dy1 .

k=1

1)
dy1 , and that [F (y1 )]1/ = exp{log F (y1 )/}, we can make
Noting that d log F (y1 ) = Ff (y
(y1 )
the substitution x = log F (y1 ) to yield

E
Thus, we have

 n


n 0
log F (Yk ) =
xex/ dx = n.

k=1


E M V (Y ) = ,

which implies that M V is an MVUE since it is an unbiased function of a complete sucient


statistic.
5

c
b. [Correction: Note that, for the given prior, the prior mean should be E{} = m1
.]
It is straightforward to see that w(|y) is of the same form as the prior with c replaced

by c nk=1 log F (yk ), and m replaced by n + m. Thus, by inspection

E{|Y } =

n

log F (Yk )
,
m+n1
k=1

which was to be shown. [Again, the necessary correction has been made.]
c. In this example, the prior and posterior distributions have the same form. The only
change is that the parameters of that distribution are updated as new data is observed.
A prior with this property is said to be a reproducing prior. The prior parameters , c and
m, can be thought of as coming from an earlier sample of size m. As n becomes large
compared to m, the importance of these prior parameters in the estimate diminishes.

Note that nk=1 log F (Yk ) behaves like nE{log F (Y1 )} for large n. Thus, with n  m, the
estimate is appropximately given by the MVUE of Part a. Altenatively, with m  n, the
estimate is approximately the prior mean, c/(m 1). Between these two extremes, there
is a balance between prior and observed information.

Exercise 23:
a. The log-likelihood function is


2

n
1 
k
yk A sin
+
log p(y|A, ) = 2
2 k=1
2

n
log(2 2 ).
2

The likelihood equations are thus:


n


k=1

and
A

n

k=1

k
yk A sin
+
2

k
sin
+ = 0,
2

k
yk A sin
+
2

k
cos
+ = 0.
2

These equations are solved by the estimates:


AM L =

yc2 + ys2 ,


yc
,
M L = tan1
ys
where

n
k
1
yk cos
yc =
n k=1
2

n/2
1

(1)k y2k ,
n k=1

and

n
k
1
yk sin
ys =
n k=1
2

n/2
1

(1)k+1 y2k1 .
n k=1

b. Appending the prior to the above problem yields MAP estimates:


M AP = M L ,
and
AM AP =

AM L +

 
2
r
n

1+

2(1+) 2
n

2
where n
2.
c. Note that, when (and the prior diuses), the MAP estimate of A does
not approach the MLE of A. However, as n , the MAP estimate does approach the
MLE.

You might also like