# Probability and Stochastic Processes

**A Friendly Introduction for Electrical and Computer Engineers
**

Second Edition

Quiz Solutions

Roy D. Yates and David J. Goodman

May 22, 2004

• The MATLAB section quizzes at the end of each chapter use programs available for

download as the archive matcode.zip. This archive has programs of general pur-

pose programs for solving probability problems as well as speciﬁc .m ﬁles associated

with examples or quizzes in the text. Also available is a manual probmatlab.pdf

describing the general purpose .m ﬁles in matcode.zip.

• We have made a substantial effort to check the solution to every quiz. Nevertheless,

there is a nonzero probability (in fact, a probability close to unity) that errors will be

found. If you ﬁnd errors or have suggestions or comments, please send email to

ryates@winlab.rutgers.edu.

When errors are found, corrected solutions will be posted at the website.

1

Quiz Solutions – Chapter 1

Quiz 1.1

In the Venn diagrams for parts (a)-(g) below, the shaded area represents the indicated

set.

M

O

T

M

O

T

M

O

T

(1) R = T

c

(2) M ∪ O (3) M ∩ O

M

O

T

M

O

T

M

O

T

(4) R ∪ M (4) R ∩ M (6) T

c

− M

Quiz 1.2

(1) A

1

= {vvv, vvd, vdv, vdd}

(2) B

1

= {dvv, dvd, ddv, ddd}

(3) A

2

= {vvv, vvd, dvv, dvd}

(4) B

2

= {vdv, vdd, ddv, ddd}

(5) A

3

= {vvv, ddd}

(6) B

3

= {vdv, dvd}

(7) A

4

= {vvv, vvd, vdv, dvv, vdd, dvd, ddv}

(8) B

4

= {ddd, ddv, dvd, vdd}

Recall that A

i

and B

i

are collectively exhaustive if A

i

∪ B

i

= S. Also, A

i

and B

i

are

mutually exclusive if A

i

∩ B

i

= φ. Since we have written down each pair A

i

and B

i

above,

we can simply check for these properties.

The pair A

1

and B

1

are mutually exclusive and collectively exhaustive. The pair A

2

and

B

2

are mutually exclusive and collectively exhaustive. The pair A

3

and B

3

are mutually

exclusive but not collectively exhaustive. The pair A

4

and B

4

are not mutually exclusive

since dvd belongs to A

4

and B

4

. However, A

4

and B

4

are collectively exhaustive.

2

Quiz 1.3

There are exactly 50 equally likely outcomes: s

51

through s

100

. Each of these outcomes

has probability 0.02.

(1) P[{s

79

}] = 0.02

(2) P[{s

100

}] = 0.02

(3) P[A] = P[{s

90

, . . . , s

100

}] = 11 ×0.02 = 0.22

(4) P[F] = P[{s

51

, . . . , s

59

}] = 9 ×0.02 = 0.18

(5) P[T ≥ 80] = P[{s

80

, . . . , s

100

}] = 21 ×0.02 = 0.42

(6) P[T < 90] = P[{s

51

, s

52

, . . . , s

89

}] = 39 ×0.02 = 0.78

(7) P[a C grade or better] = P[{s

70

, . . . , s

100

}] = 31 ×0.02 = 0.62

(8) P[student passes] = P[{s

60

, . . . , s

100

}] = 41 ×0.02 = 0.82

Quiz 1.4

We can describe this experiment by the event space consisting of the four possible

events V B, V L, DB, and DL. We represent these events in the table:

V D

L 0.35 ?

B ? ?

In a roundabout way, the problem statement tells us how to ﬁll in the table. In particular,

P [V] = 0.7 = P [V L] + P [V B] (1)

P [L] = 0.6 = P [V L] + P [DL] (2)

Since P[V L] = 0.35, we can conclude that P[V B] = 0.35 and that P[DL] = 0.6 −

0.35 = 0.25. This allows us to ﬁll in two more table entries:

V D

L 0.35 0.25

B 0.35 ?

The remaining table entry is ﬁlled in by observing that the probabilities must sum to 1.

This implies P[DB] = 0.05 and the complete table is

V D

L 0.35 0.25

B 0.35 0.05

Finding the various probabilities is now straightforward:

3

(1) P[DL] = 0.25

(2) P[D ∪ L] = P[V L] + P[DL] + P[DB] = 0.35 +0.25 +0.05 = 0.65.

(3) P[V B] = 0.35

(4) P[V ∪ L] = P[V] + P[L] − P[V L] = 0.7 +0.6 −0.35 = 0.95

(5) P[V ∪ D] = P[S] = 1

(6) P[LB] = P[LL

c

] = 0

Quiz 1.5

(1) The probability of exactly two voice calls is

P [N

V

= 2] = P [{vvd, vdv, dvv}] = 0.3 (1)

(2) The probability of at least one voice call is

P [N

V

≥ 1] = P [{vdd, dvd, ddv, vvd, vdv, dvv, vvv}] (2)

= 6(0.1) +0.2 = 0.8 (3)

An easier way to get the same answer is to observe that

P [N

V

≥ 1] = 1 − P [N

V

< 1] = 1 − P [N

V

= 0] = 1 − P [{ddd}] = 0.8 (4)

(3) The conditional probability of two voice calls followed by a data call given that there

were two voice calls is

P [{vvd} |N

V

= 2] =

P [{vvd} , N

V

= 2]

P [N

V

= 2]

=

P [{vvd}]

P [N

V

= 2]

=

0.1

0.3

=

1

3

(5)

(4) The conditional probability of two data calls followed by a voice call given there

were two voice calls is

P [{ddv} |N

V

= 2] =

P [{ddv} , N

V

= 2]

P [N

V

= 2]

= 0 (6)

The joint event of the outcome ddv and exactly two voice calls has probability zero

since there is only one voice call in the outcome ddv.

(5) The conditional probability of exactly two voice calls given at least one voice call is

P [N

V

= 2|N

v

≥ 1] =

P [N

V

= 2, N

V

≥ 1]

P [N

V

≥ 1]

=

P [N

V

= 2]

P [N

V

≥ 1]

=

0.3

0.8

=

3

8

(7)

(6) The conditional probability of at least one voice call given there were exactly two

voice calls is

P [N

V

≥ 1|N

V

= 2] =

P [N

V

≥ 1, N

V

= 2]

P [N

V

= 2]

=

P [N

V

= 2]

P [N

V

= 2]

= 1 (8)

Given that there were two voice calls, there must have been at least one voice call.

4

Quiz 1.6

In this experiment, there are four outcomes with probabilities

P[{vv}] = (0.8)

2

= 0.64 P[{vd}] = (0.8)(0.2) = 0.16

P[{dv}] = (0.2)(0.8) = 0.16 P[{dd}] = (0.2)

2

= 0.04

When checking the independence of any two events A and B, it’s wise to avoid intuition

and simply check whether P[AB] = P[A]P[B]. Using the probabilities of the outcomes,

we now can test for the independence of events.

(1) First, we calculate the probability of the joint event:

P [N

V

= 2, N

V

≥ 1] = P [N

V

= 2] = P [{vv}] = 0.64 (1)

Next, we observe that

P [N

V

≥ 1] = P [{vd, dv, vv}] = 0.96 (2)

Finally, we make the comparison

P [N

V

= 2] P [N

V

≥ 1] = (0.64)(0.96) = P [N

V

= 2, N

V

≥ 1] (3)

which shows the two events are dependent.

(2) The probability of the joint event is

P [N

V

≥ 1, C

1

= v] = P [{vd, vv}] = 0.80 (4)

From part (a), P[N

V

≥ 1] = 0.96. Further, P[C

1

= v] = 0.8 so that

P [N

V

≥ 1] P [C

1

= v] = (0.96)(0.8) = 0.768 = P [N

V

≥ 1, C

1

= v] (5)

Hence, the events are dependent.

(3) The problem statement that the calls were independent implies that the events the

second call is a voice call, {C

2

= v}, and the ﬁrst call is a data call, {C

1

= d} are

independent events. Just to be sure, we can do the calculations to check:

P [C

1

= d, C

2

= v] = P [{dv}] = 0.16 (6)

Since P[C

1

= d]P[C

2

= v] = (0.2)(0.8) = 0.16, we conﬁrm that the events are

independent. Note that this shouldn’t be surprising since we used the information that

the calls were independent in the problem statement to determine the probabilities of

the outcomes.

(4) The probability of the joint event is

P [C

2

= v, N

V

is even] = P [{vv}] = 0.64 (7)

Also, each event has probability

P [C

2

= v] = P [{dv, vv}] = 0.8, P [N

V

is even] = P [{dd, vv}] = 0.68 (8)

Thus, P[C

2

= v]P[N

V

is even] = (0.8)(0.68) = 0.544. Since P[C

2

= v, N

V

is even] =

0.544, the events are dependent.

5

Quiz 1.7

Let F

i

denote the event that that the user is found on page i . The tree for the experiment

is

¨

¨

¨

¨

¨

¨

F

1

0.8

F

c

1

0.2

¨

¨

¨

¨

¨

¨

F

2

0.8

F

c

2

0.2

¨

¨

¨

¨

¨

¨

F

3

0.8

F

c

3

0.2

The user is found unless all three paging attempts fail. Thus the probability the user is

found is

P [F] = 1 − P

¸

F

c

1

F

c

2

F

c

3

¸

= 1 −(0.2)

3

= 0.992 (1)

Quiz 1.8

(1) We can view choosing each bit in the code word as a subexperiment. Each subex-

periment has two possible outcomes: 0 and 1. Thus by the fundamental principle of

counting, there are 2 ×2 ×2 ×2 = 2

4

= 16 possible code words.

(2) An experiment that can yield all possible code words with two zeroes is to choose

which 2 bits (out of 4 bits) will be zero. The other two bits then must be ones. There

are

4

2

**= 6 ways to do this. Hence, there are six code words with exactly two zeroes.
**

For this problem, it is also possible to simply enumerate the six code words:

1100, 1010, 1001, 0101, 0110, 0011.

(3) When the ﬁrst bit must be a zero, then the ﬁrst subexperiment of choosing the ﬁrst

bit has only one outcome. For each of the next three bits, we have two choices. In

this case, there are 1 ×2 ×2 ×2 = 8 ways of choosing a code word.

(4) For the constant ratio code, we can specify a code word by choosing M of the bits to

be ones. The other N −M bits will be zeroes. The number of ways of choosing such

a code word is

N

M

. For N = 8 and M = 3, there are

8

3

= 56 code words.

Quiz 1.9

(1) In this problem, k bits received in error is the same as k failures in 100 trials. The

failure probability is = 1 − p and the success probability is 1 − = p. That is, the

probability of k bits in error and 100 −k correctly received bits is

P

¸

S

k,100−k

¸

=

100

k

k

(1 −)

100−k

(1)

6

For = 0.01,

P

¸

S

0,100

¸

= (1 −)

100

= (0.99)

100

= 0.3660 (2)

P

¸

S

1,99

¸

= 100(0.01)(0.99)

99

= 0.3700 (3)

P

¸

S

2,98

¸

= 4950(0.01)

2

(0.99)

9

8 = 0.1849 (4)

P

¸

S

3,97

¸

= 161, 700(0.01)

3

(0.99)

97

= 0.0610 (5)

(2) The probability a packet is decoded correctly is just

P [C] = P

¸

S

0,100

¸

+ P

¸

S

1,99

¸

+ P

¸

S

2,98

¸

+ P

¸

S

3,97

¸

= 0.9819 (6)

Quiz 1.10

Since the chip works only if all n transistors work, the transistors in the chip are like

devices in series. The probability that a chip works is P[C] = p

n

.

The module works if either 8 chips work or 9 chips work. Let C

k

denote the event that

exactly k chips work. Since transistor failures are independent of each other, chip failures

are also independent. Thus each P[C

k

] has the binomial probability

P [C

8

] =

9

8

(P [C])

8

(1 − P [C])

9−8

= 9p

8n

(1 − p

n

), (1)

P [C

9

] = (P [C])

9

= p

9n

. (2)

The probability a memory module works is

P [M] = P [C

8

] + P [C

9

] = p

8n

(9 −8p

n

) (3)

Quiz 1.11

R=rand(1,100);

X=(R<= 0.4) ...

+ (2*(R>0.4).*(R<=0.9)) ...

+ (3*(R>0.9));

Y=hist(X,1:3)

For a MATLAB simulation, we ﬁrst gen-

erate a vector R of 100 random numbers.

Second, we generate vector X as a func-

tion of R to represent the 3 possible out-

comes of a ﬂip. That is, X(i)=1 if ﬂip i

was heads, X(i)=2 if ﬂip i was tails, and

X(i)=3) is ﬂip i landed on the edge.

To see how this works, we note there are three cases:

• If R(i) <= 0.4, then X(i)=1.

• If 0.4 < R(i) and R(i)<=0.9, then X(i)=2.

• If 0.9 < R(i), then X(i)=3.

These three cases will have probabilities 0.4, 0.5 and 0.1. Lastly, we use the hist function

to count how many occurences of each possible value of X(i).

7

Quiz Solutions – Chapter 2

Quiz 2.1

The sample space, probabilities and corresponding grades for the experiment are

Outcome P[·] G

BB 0.36 3.0

BC 0.24 2.5

CB 0.24 2.5

CC 0.16 2

Quiz 2.2

(1) To ﬁnd c, we recall that the PMF must sum to 1. That is,

3

¸

n=1

P

N

(n) = c

1 +

1

2

+

1

3

= 1 (1)

This implies c = 6/11. Now that we have found c, the remaining parts are straight-

forward.

(2) P[N = 1] = P

N

(1) = c = 6/11

(3) P[N ≥ 2] = P

N

(2) + P

N

(3) = c/2 +c/3 = 5/11

(4) P[N > 3] =

¸

∞

n=4

P

N

(n) = 0

Quiz 2.3

Decoding each transmitted bit is an independent trial where we call a bit error a “suc-

cess.” Each bit is in error, that is, the trial is a success, with probability p. Now we can

interpret each experiment in the generic context of independent trials.

(1) The random variable X is the number of trials up to and including the ﬁrst success.

Similar to Example 2.11, X has the geometric PMF

P

X

(x) =

¸

p(1 − p)

x−1

x = 1, 2, . . .

0 otherwise

(1)

(2) If p = 0.1, then the probability exactly 10 bits are sent is

P [X = 10] = P

X

(10) = (0.1)(0.9)

9

= 0.0387 (2)

8

The probability that at least 10 bits are sent is P[X ≥ 10] =

¸

∞

x=10

P

X

(x). This

sum is not too hard to calculate. However, its even easier to observe that X ≥ 10 if

the ﬁrst 10 bits are transmitted correctly. That is,

P [X ≥ 10] = P [ﬁrst 10 bits are correct] = (1 − p)

10

(3)

For p = 0.1, P[X ≥ 10] = 0.9

10

= 0.3487.

(3) The random variable Y is the number of successes in 100 independent trials. Just as

in Example 2.13, Y has the binomial PMF

P

Y

(y) =

100

y

p

y

(1 − p)

100−y

(4)

If p = 0.01, the probability of exactly 2 errors is

P [Y = 2] = P

Y

(2) =

100

2

(0.01)

2

(0.99)

98

= 0.1849 (5)

(4) The probability of no more than 2 errors is

P [Y ≤ 2] = P

Y

(0) + P

Y

(1) + P

Y

(2) (6)

= (0.99)

100

+100(0.01)(0.99)

99

+

100

2

(0.01)

2

(0.99)

98

(7)

= 0.9207 (8)

(5) Random variable Z is the number of trials up to and including the third success. Thus

Z has the Pascal PMF (see Example 2.15)

P

Z

(z) =

z −1

2

p

3

(1 − p)

z−3

(9)

Note that P

Z

(z) > 0 for z = 3, 4, 5, . . ..

(6) If p = 0.25, the probability that the third error occurs on bit 12 is

P

Z

(12) =

11

2

(0.25)

3

(0.75)

9

= 0.0645 (10)

Quiz 2.4

Each of these probabilities can be read off the CDF F

Y

(y). However, we must keep in

mind that when F

Y

(y) has a discontinuity at y

0

, F

Y

(y) takes the upper value F

Y

(y

+

0

).

(1) P[Y < 1] = F

Y

(1

−

) = 0

9

(2) P[Y ≤ 1] = F

Y

(1) = 0.6

(3) P[Y > 2] = 1 − P[Y ≤ 2] = 1 − F

Y

(2) = 1 −0.8 = 0.2

(4) P[Y ≥ 2] = 1 − P[Y < 2] = 1 − F

Y

(2

−

) = 1 −0.6 = 0.4

(5) P[Y = 1] = P[Y ≤ 1] − P[Y < 1] = F

Y

(1

+

) − F

Y

(1

−

) = 0.6

(6) P[Y = 3] = P[Y ≤ 3] − P[Y < 3] = F

Y

(3

+

) − F

Y

(3

−

) = 0.8 −0.8 = 0

Quiz 2.5

(1) With probability 0.7, a call is a voice call and C = 25. Otherwise, with probability

0.3, we have a data call and C = 40. This corresponds to the PMF

P

C

(c) =

⎧

⎨

⎩

0.7 c = 25

0.3 c = 40

0 otherwise

(1)

(2) The expected value of C is

E [C] = 25(0.7) +40(0.3) = 29.5 cents (2)

Quiz 2.6

(1) As a function of N, the cost T is

T = 25N +40(3 − N) = 120 −15N (1)

(2) To ﬁnd the PMF of T, we can draw the following tree:

¨

¨

¨

¨

¨

¨

¨

N=0

0.1

r

r

r

r

r

r

r

N=3

0.3

$

$

$

$

$

$

$N=1 0.3

N=2 0.3

•T=120

•T=105

•T=90

•T=75

From the tree, we can write down the PMF of T:

P

T

(t ) =

⎧

⎨

⎩

0.3 t = 75, 90, 105

0.1 t = 120

0 otherwise

(2)

From the PMF P

T

(t ), the expected value of T is

E [T] = 75P

T

(75) +90P

T

(90) +105P

T

(105) +120P

T

(120) (3)

= (75 +90 +105)(0.3) +120(0.1) = 62 (4)

10

Quiz 2.7

(1) Using Deﬁnition 2.14, the expected number of applications is

E [A] =

4

¸

a=1

aP

A

(a) = 1(0.4) +2(0.3) +3(0.2) +4(0.1) = 2 (1)

(2) The number of memory chips is M = g(A) where

g(A) =

⎧

⎨

⎩

4 A = 1, 2

6 A = 3

8 A = 4

(2)

(3) By Theorem 2.10, the expected number of memory chips is

E [M] =

4

¸

a=1

g(A)P

A

(a) = 4(0.4) +4(0.3) +6(0.2) +8(0.1) = 4.8 (3)

Since E[A] = 2, g(E[A]) = g(2) = 4. However, E[M] = 4.8 = g(E[A]). The two

quantities are different because g(A) is not of the form αA +β.

Quiz 2.8

The PMF P

N

(n) allows to calculate each of the desired quantities.

(1) The expected value of N is

E [N] =

2

¸

n=0

nP

N

(n) = 0(0.1) +1(0.4) +2(0.5) = 1.4 (1)

(2) The second moment of N is

E

¸

N

2

¸

=

2

¸

n=0

n

2

P

N

(n) = 0

2

(0.1) +1

2

(0.4) +2

2

(0.5) = 2.4 (2)

(3) The variance of N is

Var[N] = E

¸

N

2

¸

−(E [N])

2

= 2.4 −(1.4)

2

= 0.44 (3)

(4) The standard deviation is σ

N

=

√

Var[N] =

√

0.44 = 0.663.

11

Quiz 2.9

(1) From the problem statement, we learn that the conditional PMF of N given the event

I is

P

N|I

(n) =

¸

0.02 n = 1, 2, . . . , 50

0 otherwise

(1)

(2) Also from the problem statement, the conditional PMF of N given the event T is

P

N|T

(n) =

¸

0.2 n = 1, 2, 3, 4, 5

0 otherwise

(2)

(3) The problem statement tells us that P[T] = 1 − P[I ] = 3/4. From Theorem 1.10

(the law of total probability), we ﬁnd the PMF of N is

P

N

(n) = P

N|T

(n) P [T] + P

N|I

(n) P [I ] (3)

=

⎧

⎨

⎩

0.2(0.75) +0.02(0.25) n = 1, 2, 3, 4, 5

0(0.75) +0.02(0.25) n = 6, 7, . . . , 50

0 otherwise

(4)

=

⎧

⎨

⎩

0.155 n = 1, 2, 3, 4, 5

0.005 n = 6, 7, . . . , 50

0 otherwise

(5)

(4) First we ﬁnd

P [N ≤ 10] =

10

¸

n=1

P

N

(n) = (0.155)(5) +(0.005)(5) = 0.80 (6)

By Theorem 2.17, the conditional PMF of N given N ≤ 10 is

P

N|N≤10

(n) =

¸

P

N

(n)

P[N≤10]

n ≤ 10

0 otherwise

(7)

=

⎧

⎨

⎩

0.155/0.8 n = 1, 2, 3, 4, 5

0.005/0.8 n = 6, 7, 8, 9, 10

0 otherwise

(8)

=

⎧

⎨

⎩

0.19375 n = 1, 2, 3, 4, 5

0.00625 n = 6, 7, 8, 9, 10

0 otherwise

(9)

(5) Once we have the conditional PMF, calculating conditional expectations is easy.

E [N|N ≤ 10] =

¸

n

nP

N|N≤10

(n) (10)

=

5

¸

n=1

n(0.19375) +

10

¸

n=6

n(0.00625) (11)

= 3.15625 (12)

12

0 50 100

0

2

4

6

8

10

0 500 1000

0

2

4

6

8

10

(a) samplemean(100) (b) samplemean(1000)

Figure 1: Two examples of the output of samplemean(k)

(6) To ﬁnd the conditional variance, we ﬁrst ﬁnd the conditional second moment

E

¸

N

2

|N ≤ 10

¸

=

¸

n

n

2

P

N|N≤10

(n) (13)

=

5

¸

n=1

n

2

(0.19375) +

10

¸

n=6

n

2

(0.00625) (14)

= 55(0.19375) +330(0.00625) = 12.71875 (15)

The conditional variance is

Var[N|N ≤ 10] = E

¸

N

2

|N ≤ 10

¸

−(E [N|N ≤ 10])

2

(16)

= 12.71875 −(3.15625)

2

= 2.75684 (17)

Quiz 2.10

The function samplemean(k) generates and plots ﬁve m

n

sequences for n = 1, 2, . . . , k.

The i th column M(:,i) of M holds a sequence m

1

, m

2

, . . . , m

k

.

function M=samplemean(k);

K=(1:k)’;

M=zeros(k,5);

for i=1:5,

X=duniformrv(0,10,k);

M(:,i)=cumsum(X)./K;

end;

plot(K,M);

Examples of the function calls (a) samplemean(100) and (b) samplemean(1000)

are shown in Figure 1. Each time samplemean(k) is called produces a random output.

What is observed in these ﬁgures is that for small n, m

n

is fairly random but as n gets

13

large, m

n

gets close to E[X] = 5. Although each sequence m

1

, m

2

, . . . that we generate is

random, the sequences always converges to E[X]. This random convergence is analyzed

in Chapter 7.

14

Quiz Solutions – Chapter 3

Quiz 3.1

The CDF of Y is

0 2 4

0

0.5

1

y

F

Y

(

y

)

F

Y

(y) =

⎧

⎨

⎩

0 y < 0

y/4 0 ≤ y ≤ 4

1 y > 4

(1)

From the CDF F

Y

(y), we can calculate the probabilities:

(1) P[Y ≤ −1] = F

Y

(−1) = 0

(2) P[Y ≤ 1] = F

Y

(1) = 1/4

(3) P[2 < Y ≤ 3] = F

Y

(3) − F

Y

(2) = 3/4 −2/4 = 1/4

(4) P[Y > 1.5] = 1 − P[Y ≤ 1.5] = 1 − F

Y

(1.5) = 1 −(1.5)/4 = 5/8

Quiz 3.2

(1) First we will ﬁnd the constant c and then we will sketch the PDF. To ﬁnd c, we use

the fact that

∞

−∞

f

X

(x) dx = 1. We will evaluate this integral using integration by

parts:

∞

−∞

f

X

(x) dx =

∞

0

cxe

−x/2

dx (1)

= −2cxe

−x/2

∞

0

. .. .

=0

+

∞

0

2ce

−x/2

dx (2)

= −4ce

−x/2

∞

0

= 4c (3)

Thus c = 1/4 and X has the Erlang (n = 2, λ = 1/2) PDF

0 5 10 15

0

0.1

0.2

x

f

X

(

x

)

f

X

(x) =

¸

(x/4)e

−x/2

x ≥ 0

0 otherwise

(4)

15

(2) To ﬁnd the CDF F

X

(x), we ﬁrst note X is a nonnegative random variable so that

F

X

(x) = 0 for all x < 0. For x ≥ 0,

F

X

(x) =

x

0

f

X

(y) dy =

x

0

y

4

e

−y/2

dy (5)

= −

y

2

e

−y/2

x

0

−

x

0

−

1

2

e

−y/2

dy (6)

= 1 −

x

2

e

−x/2

−e

−x/2

(7)

The complete expression for the CDF is

0 5 10 15

0

0.5

1

x

F

X

(

x

)

F

X

(x) =

¸

1 −

x

2

+1

e

−x/2

x ≥ 0

0 otherwise

(8)

(3) From the CDF F

X

(x),

P [0 ≤ X ≤ 4] = F

X

(4) − F

X

(0) = 1 −3e

−2

. (9)

(4) Similarly,

P [−2 ≤ X ≤ 2] = F

X

(2) − F

X

(−2) = 1 −3e

−1

. (10)

Quiz 3.3

The PDF of Y is

−2 0 2

0

1

2

3

y

f

Y

(

y

)

f

Y

(y) =

¸

3y

2

/2 −1 ≤ y ≤ 1,

0 otherwise.

(1)

(1) The expected value of Y is

E [Y] =

∞

−∞

y f

Y

(y) dy =

1

−1

(3/2)y

3

dy = (3/8)y

4

1

−1

= 0. (2)

Note that the above calculation wasn’t really necessary because E[Y] = 0 whenever

the PDF f

Y

(y) is an even function (i.e., f

Y

(y) = f

Y

(−y)).

(2) The second moment of Y is

E

¸

Y

2

¸

=

∞

−∞

y

2

f

Y

(y) dy =

1

−1

(3/2)y

4

dy = (3/10)y

5

1

−1

= 3/5. (3)

16

(3) The variance of Y is

Var[Y] = E

¸

Y

2

¸

−(E [Y])

2

= 3/5. (4)

(4) The standard deviation of Y is σ

Y

=

√

Var[Y] =

√

3/5.

Quiz 3.4

(1) When X is an exponential (λ) random variable, E[X] = 1/λ and Var[X] = 1/λ

2

.

Since E[X] = 3 and Var[X] = 9, we must have λ = 1/3. The PDF of X is

f

X

(x) =

¸

(1/3)e

−x/3

x ≥ 0,

0 otherwise.

(1)

(2) We know X is a uniform (a, b) random variable. To ﬁnd a and b, we apply Theo-

rem 3.6 to write

E [X] =

a +b

2

= 3 Var[X] =

(b −a)

2

12

= 9. (2)

This implies

a +b = 6, b −a = ±6

√

3. (3)

The only valid solution with a < b is

a = 3 −3

√

3, b = 3 +3

√

3. (4)

The complete expression for the PDF of X is

f

X

(x) =

¸

1/(6

√

3) 3 −3

√

3 ≤ x < 3 +3

√

3,

0 otherwise.

(5)

Quiz 3.5

Each of the requested probabilities can be calculated using (z) function and Table 3.1

or Q(z) and Table 3.2. We start with the sketches.

(1) The PDFs of X and Y are shown below. The fact that Y has twice the standard

deviation of X is reﬂected in the greater spread of f

Y

(y). However, it is important

to remember that as the standard deviation increases, the peak value of the Gaussian

PDF goes down.

−5 0 5

0

0.2

0.4

x y

f

X

(

x

)

f

Y

(

y

)

← f

X

(x)

← f

Y

(y)

17

(2) Since X is Gaussian (0, 1),

P [−1 < X ≤ 1] = F

X

(1) − F

X

(−1) (1)

= (1) −(−1) = 2(1) −1 = 0.6826. (2)

(3) Since Y is Gaussian (0, 2),

P [−1 < Y ≤ 1] = F

Y

(1) − F

Y

(−1) (3)

=

1

σ

Y

−

−1

σ

Y

= 2

1

2

−1 = 0.383. (4)

(4) Again, since X is Gaussian (0, 1), P[X > 3.5] = Q(3.5) = 2.33 ×10

−4

.

(5) Since Y is Gaussian (0, 2), P[Y > 3.5] = Q(

3.5

2

) = Q(1.75) = 1 − (1.75) =

0.0401.

Quiz 3.6

The CDF of X is

−2 0 2

0

0.5

1

x

F

X

(

x

)

F

X

(x) =

⎧

⎨

⎩

0 x < −1,

(x +1)/4 −1 ≤ x < 1,

1 x ≥ 1.

(1)

The following probabilities can be read directly from the CDF:

(1) P[X ≤ 1] = F

X

(1) = 1.

(2) P[X < 1] = F

X

(1

−

) = 1/2.

(3) P[X = 1] = F

X

(1

+

) − F

X

(1

−

) = 1 −1/2 = 1/2.

(4) We ﬁnd the PDF f

Y

(y) by taking the derivative of F

Y

(y). The resulting PDF is

−2 0 2

0

0.5

x

f

X

(

x

)

0.5

f

X

(x) =

⎧

⎨

⎩

1/4 −1 ≤ x < 1,

(1/2)δ(x −1) x = 1,

0 otherwise.

(2)

Quiz 3.7

18

(1) Since X is always nonnegative, F

X

(x) = 0 for x < 0. Also, F

X

(x) = 1 for x ≥ 2

since its always true that x ≤ 2. Lastly, for 0 ≤ x ≤ 2,

F

X

(x) =

x

−∞

f

X

(y) dy =

x

0

(1 − y/2) dy = x − x

2

/4. (1)

The complete CDF of X is

−1 0 1 2 3

0

0.5

1

x

F

X

(

x

)

F

X

(x) =

⎧

⎨

⎩

0 x < 0,

x − x

2

/4 0 ≤ x ≤ 2,

1 x > 2.

(2)

(2) The probability that Y = 1 is

P [Y = 1] = P [X ≥ 1] = 1 − F

X

(1) = 1 −3/4 = 1/4. (3)

(3) Since X is nonnegative, Y is also nonnegative. Thus F

Y

(y) = 0 for y < 0. Also,

because Y ≤ 1, F

Y

(y) = 1 for all y ≥ 1. Finally, for 0 < y < 1,

F

Y

(y) = P [Y ≤ y] = P [X ≤ y] = F

X

(y) . (4)

Using the CDF F

X

(x), the complete expression for the CDF of Y is

−1 0 1 2 3

0

0.5

1

y

F

Y

(

y

)

F

Y

(y) =

⎧

⎨

⎩

0 y < 0,

y − y

2

/4 0 ≤ y < 1,

1 y ≥ 1.

(5)

As expected, we see that the jump in F

Y

(y) at y = 1 is exactly equal to P[Y = 1].

(4) By taking the derivative of F

Y

(y), we obtain the PDF f

Y

(y). Note that when y < 0

or y > 1, the PDF is zero.

−1 0 1 2 3

0

0.5

1

1.5

y

f

Y

(

y

)

0.25

f

Y

(y) =

¸

1 − y/2 +(1/4)δ(y −1) 0 ≤ y ≤ 1

0 otherwise

(6)

Quiz 3.8

(1) P[Y ≤ 6] =

6

−∞

f

Y

(y) dy =

6

0

(1/10) dy = 0.6 .

19

(2) From Deﬁnition 3.15, the conditional PDF of Y given Y ≤ 6 is

f

Y|Y≤6

(y) =

¸

f

Y

(y)

P[Y≤6]

y ≤ 6,

0 otherwise,

=

¸

1/6 0 ≤ y ≤ 6,

0 otherwise.

(1)

(3) The probability Y > 8 is

P [Y > 8] =

10

8

1

10

dy = 0.2 . (2)

(4) From Deﬁnition 3.15, the conditional PDF of Y given Y > 8 is

f

Y|Y>8

(y) =

¸

f

Y

(y)

P[Y>8]

y > 8,

0 otherwise,

=

¸

1/2 8 < y ≤ 10,

0 otherwise.

(3)

(5) From the conditional PDF f

Y|Y≤6

(y), we can calculate the conditional expectation

E [Y|Y ≤ 6] =

∞

−∞

y f

Y|Y≤6

(y) dy =

6

0

y

6

dy = 3. (4)

(6) From the conditional PDF f

Y|Y>8

(y), we can calculate the conditional expectation

E [Y|Y > 8] =

∞

−∞

y f

Y|Y>8

(y) dy =

10

8

y

2

dy = 9. (5)

Quiz 3.9

A natural way to produce random variables with PDF f

T|T>2

(t ) is to generate samples

of T with PDF f

T

(t ) and then to discard those samples which fail to satisfy the condition

T > 2. Here is a MATLAB function that uses this method:

function t=t2rv(m)

i=0;lambda=1/3;

t=zeros(m,1);

while (i<m),

x=exponentialrv(lambda,1);

if (x>2)

t(i+1)=x;

i=i+1;

end

end

A second method exploits the fact that if T is an exponential (λ) random variable, then

T

= T +2 has PDF f

T

(t ) = f

T|T>2

(t ). In this case the command

t=2.0+exponentialrv(1/3,m)

generates the vector t.

20

Quiz Solutions – Chapter 4

Quiz 4.1

Each value of the joint CDF can be found by considering the corresponding probability.

(1) F

X,Y

(−∞, 2) = P[X ≤ −∞, Y ≤ 2] ≤ P[X ≤ −∞] = 0 since X cannot take on

the value −∞.

(2) F

X,Y

(∞, ∞) = P[X ≤ ∞, Y ≤ ∞] = 1. This result is given in Theorem 4.1.

(3) F

X,Y

(∞, y) = P[X ≤ ∞, Y ≤ y] = P[Y ≤ y] = F

Y

(y).

(4) F

X,Y

(∞, −∞) = P[X ≤ ∞, Y ≤ −∞] = 0 since Y cannot take on the value −∞.

Quiz 4.2

From the joint PMF of Q and G given in the table, we can calculate the requested

probabilities by summing the PMF over those values of Q and G that correspond to the

event.

(1) The probability that Q = 0 is

P [Q = 0] = P

Q,G

(0, 0) + P

Q,G

(0, 1) + P

Q,G

(0, 2) + P

Q,G

(0, 3) (1)

= 0.06 +0.18 +0.24 +0.12 = 0.6 (2)

(2) The probability that Q = G is

P [Q = G] = P

Q,G

(0, 0) + P

Q,G

(1, 1) = 0.18 (3)

(3) The probability that G > 1 is

P [G > 1] =

3

¸

g=2

1

¸

q=0

P

Q,G

(q, g) (4)

= 0.24 +0.16 +0.12 +0.08 = 0.6 (5)

(4) The probability that G > Q is

P [G > Q] =

1

¸

q=0

3

¸

g=q+1

P

Q,G

(q, g) (6)

= 0.18 +0.24 +0.12 +0.16 +0.08 = 0.78 (7)

21

Quiz 4.3

By Theorem 4.3, the marginal PMF of H is

P

H

(h) =

¸

b=0,2,4

P

H,B

(h, b) (1)

For each value of h, this corresponds to calculating the row sum across the table of the joint

PMF. Similarly, the marginal PMF of B is

P

B

(b) =

1

¸

h=−1

P

H,B

(h, b) (2)

For each value of b, this corresponds to the column sum down the table of the joint PMF.

The easiest way to calculate these marginal PMFs is to simply sum each row and column:

P

H,B

(h, b) b = 0 b = 2 b = 4 P

H

(h)

h = −1 0 0.4 0.2 0.6

h = 0 0.1 0 0.1 0.2

h = 1 0.1 0.1 0 0.2

P

B

(b) 0.2 0.5 0.3

(3)

Quiz 4.4

To ﬁnd the constant c, we apply

∞

−∞

∞

−∞

f

X,Y

(x, y) dx dy = 1. Speciﬁcally,

∞

−∞

∞

−∞

f

X,Y

(x, y) dx dy =

2

0

1

0

cxy dx dy (1)

= c

2

0

y

x

2

/2

1

0

dy (2)

= (c/2)

2

0

y dy = (c/4)y

2

2

0

= c (3)

Thus c = 1. To calculate P[A], we write

P [A] =

A

f

X,Y

(x, y) dx dy (4)

To integrate over A, we convert to polar coordinates using the substitutions x = r cos θ,

y = r sin θ and dx dy = r dr dθ, yielding

Y

X

1

1

2

A

P [A] =

π/2

0

1

0

r

2

sin θ cos θ r dr dθ (5)

=

1

0

r

3

dr

π/2

0

sin θ cos θ dθ

(6)

=

r

4

/4

1

0

⎛

⎝

sin

2

θ

2

π/2

0

⎞

⎠

= 1/8 (7)

22

Quiz 4.5

By Theorem 4.8, the marginal PDF of X is

f

X

(x) =

∞

−∞

f

X,Y

(x, y) dy (1)

For x < 0 or x > 1, f

X

(x) = 0. For 0 ≤ x ≤ 1,

f

X

(x) =

6

5

1

0

(x + y

2

) dy =

6

5

xy + y

3

/3

y=1

y=0

=

6

5

(x +1/3) =

6x +2

5

(2)

The complete expression for the PDf of X is

f

X

(x) =

¸

(6x +2)/5 0 ≤ x ≤ 1

0 otherwise

(3)

By the same method we obtain the marginal PDF for Y. For 0 ≤ y ≤ 1,

f

Y

(y) =

∞

−∞

f

X,Y

(x, y) dy (4)

=

6

5

1

0

(x + y

2

) dx =

6

5

x

2

/2 + xy

2

x=1

x=0

=

6

5

(1/2 + y

2

) =

3 +6y

2

5

(5)

Since f

Y

(y) = 0 for y < 0 or y > 1, the complete expression for the PDF of Y is

f

Y

(y) =

¸

(3 +6y

2

)/5 0 ≤ y ≤ 1

0 otherwise

(6)

Quiz 4.6

(A) The time required for the transfer is T = L/B. For each pair of values of L and B,

we can calculate the time T needed for the transfer. We can write these down on the

table for the joint PMF of L and B as follows:

P

L,B

(l, b) b = 14, 400 b = 21, 600 b = 28, 800

l = 518, 400 0.20 (T=36) 0.10 (T=24) 0.05 (T=18)

l = 2, 592, 000 0.05 (T=180) 0.10 (T=120) 0.20 (T=90)

l = 7, 776, 000 0.00 (T=540) 0.10 (T=360) 0.20 (T=270)

From the table, writing down the PMF of T is straightforward.

P

T

(t ) =

⎧

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎨

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎪

⎩

0.05 t = 18

0.1 t = 24

0.2 t = 36, 90

0.1 t = 120

0.05 t = 180

0.2 t = 270

0.1 t = 360

0 otherwise

(1)

23

(B) First, we observe that since 0 ≤ X ≤ 1 and 0 ≤ Y ≤ 1, W = XY satisﬁes

0 ≤ W ≤ 1. Thus f

W

(0) = 0 and f

W

(1) = 1. For 0 < w < 1, we calculate the

CDF F

W

(w) = P[W ≤ w]. As shown below, integrating over the region W ≤ w

is fairly complex. The calculus is simpler if we integrate over the region XY > w.

Speciﬁcally,

Y

X

1

1

XY > w

w

w

XY = w

F

W

(w) = 1 − P [XY > w] (2)

= 1 −

1

w

1

w/x

dy dx (3)

= 1 −

1

w

(1 −w/x) dx (4)

= 1 −

x −wln x|

x=1

x=w

(5)

= 1 −(1 −w +wln w) = w −wln w (6)

The complete expression for the CDF is

F

W

(w) =

⎧

⎨

⎩

0 w < 0

w −wln w 0 ≤ w ≤ 1

1 w > 1

(7)

By taking the derivative of the CDF, we ﬁnd the PDF is

f

W

(w) =

d F

W

(w)

dw

=

⎧

⎨

⎩

0 w < 0

−ln w 0 ≤ w ≤ 1

0 w > 1

(8)

Quiz 4.7

(A) It is helpful to ﬁrst make a table that includes the marginal PMFs.

P

L,T

(l, t ) t = 40 t = 60 P

L

(l)

l = 1 0.15 0.1 0.25

l = 2 0.3 0.2 0.5

l = 3 0.15 0.1 0.25

P

T

(t ) 0.6 0.4

(1) The expected value of L is

E [L] = 1(0.25) +2(0.5) +3(0.25) = 2. (1)

Since the second moment of L is

E

¸

L

2

¸

= 1

2

(0.25) +2

2

(0.5) +3

2

(0.25) = 4.5, (2)

the variance of L is

Var [L] = E

¸

L

2

¸

−(E [L])

2

= 0.5. (3)

24

(2) The expected value of T is

E [T] = 40(0.6) +60(0.4) = 48. (4)

The second moment of T is

E

¸

T

2

¸

= 40

2

(0.6) +60

2

(0.4) = 2400. (5)

Thus

Var[T] = E

¸

T

2

¸

−(E [T])

2

= 2400 −48

2

= 96. (6)

(3) The correlation is

E [LT] =

¸

t =40,60

3

¸

l=1

lt P

LT

(lt ) (7)

= 1(40)(0.15) +2(40)(0.3) +3(40)(0.15) (8)

+1(60)(0.1) +2(60)(0.2) +3(60)(0.1) (9)

= 96 (10)

(4) From Theorem 4.16(a), the covariance of L and T is

Cov [L, T] = E [LT] − E [L] E [T] = 96 −2(48) = 0 (11)

(5) Since Cov[L, T] = 0, the correlation coefﬁcient is ρ

L,T

= 0.

(B) As in the discrete case, the calculations become easier if we ﬁrst calculate the marginal

PDFs f

X

(x) and f

Y

(y). For 0 ≤ x ≤ 1,

f

X

(x) =

∞

−∞

f

X,Y

(x, y) dy =

2

0

xy dy =

1

2

xy

2

y=2

y=0

= 2x (12)

Similarly, for 0 ≤ y ≤ 2,

f

Y

(y) =

∞

−∞

f

X,Y

(x, y) dx =

2

0

xy dx =

1

2

x

2

y

x=1

x=0

=

y

2

(13)

The complete expressions for the marginal PDFs are

f

X

(x) =

¸

2x 0 ≤ x ≤ 1

0 otherwise

f

Y

(y) =

¸

y/2 0 ≤ y ≤ 2

0 otherwise

(14)

From the marginal PDFs, it is straightforward to calculate the various expectations.

25

(1) The ﬁrst and second moments of X are

E [X] =

∞

−∞

x f

X

(x) dx =

1

0

2x

2

dx =

2

3

(15)

E

¸

X

2

¸

=

∞

−∞

x

2

f

X

(x) dx =

1

0

2x

3

dx =

1

2

(16)

(17)

The variance of X is Var[X] = E[X

2

] −(E[X])

2

= 1/18.

(2) The ﬁrst and second moments of Y are

E [Y] =

∞

−∞

y f

Y

(y) dy =

2

0

1

2

y

2

dy =

4

3

(18)

E

¸

Y

2

¸

=

∞

−∞

y

2

f

Y

(y) dy =

2

0

1

2

y

3

dy = 2 (19)

The variance of Y is Var[Y] = E[Y

2

] −(E[Y])

2

= 2 −16/9 = 2/9.

(3) The correlation of X and Y is

E [XY] =

∞

−∞

∞

−∞

xy f

X,Y

(x, y) dx, dy (20)

=

1

0

2

0

x

2

y

2

dx, dy =

x

3

3

1

0

y

3

3

2

0

=

8

9

(21)

(4) The covariance of X and Y is

Cov [X, Y] = E [XY] − E [X] E [Y] =

8

9

−

2

3

4

3

= 0. (22)

(5) Since Cov[X, Y] = 0, the correlation coefﬁcient is ρ

X,Y

= 0.

Quiz 4.8

(A) Since the event V > 80 occurs only for the pairs (L, T) = (2, 60), (L, T) = (3, 40)

and (L, T) = (3, 60),

P [A] = P [V > 80] = P

L,T

(2, 60) + P

L,T

(3, 40) + P

L,T

(3, 60) = 0.45 (1)

By Deﬁnition 4.9,

P

L,T| A

(l, t ) =

¸

P

L,T

(l,t )

P[A]

lt > 80

0 otherwise

(2)

26

We can represent this conditional PMF in the following table:

P

L,T| A

(l, t ) t = 40 t = 60

l = 1 0 0

l = 2 0 4/9

l = 3 1/3 2/9

The conditional expectation of V can be found from the conditional PMF.

E [V| A] =

¸

l

¸

t

lt P

L,T| A

(l, t ) (3)

= (2 · 60)

4

9

+(3 · 40)

1

3

+(3 · 60)

2

9

= 133

1

3

(4)

For the conditional variance Var[V| A], we ﬁrst ﬁnd the conditional second moment

E

¸

V

2

| A

¸

=

¸

l

¸

t

(lt )

2

P

L,T| A

(l, t ) (5)

= (2 · 60)

2

4

9

+(3 · 40)

2

1

3

+(3 · 60)

2

2

9

= 18, 400 (6)

It follows that

Var [V| A] = E

¸

V

2

| A

¸

−(E [V| A])

2

= 622

2

9

(7)

(B) For continuous random variables X and Y, we ﬁrst calculate the probability of the

conditioning event.

P [B] =

B

f

X,Y

(x, y) dx dy =

60

40

3

80/y

xy

4000

dx dy (8)

=

60

40

y

4000

x

2

2

3

80/y

dy (9)

=

60

40

y

4000

9

2

−

3200

y

2

dy (10)

=

9

8

−

4

5

ln

3

2

≈ 0.801 (11)

The conditional PDF of X and Y is

f

X,Y|B

(x, y) =

¸

f

X,Y

(x, y) /P [B] (x, y) ∈ B

0 otherwise

(12)

=

¸

Kxy 40 ≤ y ≤ 60, 80/y ≤ x ≤ 3

0 otherwise

(13)

27

where K = (4000P[B])

−1

. The conditional expectation of W given event B is

E [W|B] =

∞

−∞

∞

−∞

xy f

X,Y|B

(x, y) dx dy (14)

=

60

40

3

80/y

Kx

2

y

2

dx dy (15)

= (K/3)

60

40

y

2

x

3

x=3

x=80/y

dy (16)

= (K/3)

60

40

27y

2

−80

3

/y

dy (17)

= (K/3)

9y

3

−80

3

ln y

60

40

≈ 120.78 (18)

The conditional second moment of K given B is

E

¸

W

2

|B

¸

=

∞

−∞

∞

−∞

(xy)

2

f

X,Y|B

(x, y) dx dy (19)

=

60

40

3

80/y

Kx

3

y

3

dx dy (20)

= (K/4)

60

40

y

3

x

4

x=3

x=80/y

dy (21)

= (K/4)

60

40

81y

3

−80

4

/y

dy (22)

= (K/4)

(81/4)y

4

−80

4

ln y

60

40

≈ 16, 116.10 (23)

It follows that the conditional variance of W given B is

Var [W|B] = E

¸

W

2

|B

¸

−(E [W|B])

2

≈ 1528.30 (24)

Quiz 4.9

(A) (1) The joint PMF of A and B can be found from the marginal and conditional

PMFs via P

A,B

(a, b) = P

B| A

(b|a)P

A

(a). Incorporating the information from

the given conditional PMFs can be confusing, however. Consequently, we can

note that A has range S

A

= {0, 2} and B has range S

B

= {0, 1}. A table of the

joint PMF will include all four possible combinations of A and B. The general

form of the table is

P

A,B

(a, b) b = 0 b = 1

a = 0 P

B| A

(0|0)P

A

(0) P

B| A

(1|0)P

A

(0)

a = 2 P

B| A

(0|2)P

A

(2) P

B| A

(1|2)P

A

(2)

28

Substituting values from P

B| A

(b|a) and P

A

(a), we have

P

A,B

(a, b) b = 0 b = 1

a = 0 (0.8)(0.4) (0.2)(0.4)

a = 2 (0.5)(0.6) (0.5)(0.6)

or

P

A,B

(a, b) b = 0 b = 1

a = 0 0.32 0.08

a = 2 0.3 0.3

(2) Given the conditional PMF P

B| A

(b|2), it is easy to calculate the conditional

expectation

E [B| A = 2] =

1

¸

b=0

bP

B| A

(b|2) = (0)(0.5) +(1)(0.5) = 0.5 (1)

(3) From the joint PMF P

A,B

(a, b), we can calculate the the conditional PMF

P

A|B

(a|0) =

P

A,B

(a, 0)

P

B

(0)

=

⎧

⎨

⎩

0.32/0.62 a = 0

0.3/0.62 a = 2

0 otherwise

(2)

=

⎧

⎨

⎩

16/31 a = 0

15/31 a = 2

0 otherwise

(3)

(4) We can calculate the conditional variance Var[A|B = 0] using the conditional

PMF P

A|B

(a|0). First we calculate the conditional expected value

E [A|B = 0] =

¸

a

aP

A|B

(a|0) = 0(16/31) +2(15/31) = 30/31 (4)

The conditional second moment is

E

¸

A

2

|B = 0

¸

=

¸

a

a

2

P

A|B

(a|0) = 0

2

(16/31) +2

2

(15/31) = 60/31 (5)

The conditional variance is then

Var[A|B = 0] = E

¸

A

2

|B = 0

¸

−(E [A|B = 0])

2

=

960

961

(6)

(B) (1) The joint PDF of X and Y is

f

X,Y

(x, y) = f

Y|X

(y|x) f

X

(x) =

¸

6y 0 ≤ y ≤ x, 0 ≤ x ≤ 1

0 otherwise

(7)

(2) From the given conditional PDF f

Y|X

(y|x),

f

Y|X

(y|1/2) =

¸

8y 0 ≤ y ≤ 1/2

0 otherwise

(8)

29

(3) The conditional PDF of Y given X = 1/2 is f

X|Y

(x|1/2) = f

X,Y

(x, 1/2)/f

Y

(1/2).

To ﬁnd f

Y

(1/2), we integrate the joint PDF.

f

Y

(1/2) =

∞

−∞

f

X,1/2

( ) dx =

1

1/2

6(1/2) dx = 3/2 (9)

Thus, for 1/2 ≤ x ≤ 1,

f

X|Y

(x|1/2) =

f

X,Y

(x, 1/2)

f

Y

(1/2)

=

6(1/2)

3/2

= 2 (10)

(4) From the pervious part, we see that given Y = 1/2, the conditional PDF of X

is uniform (1/2, 1). Thus, by the deﬁnition of the uniform (a, b) PDF,

Var [X|Y = 1/2] =

(1 −1/2)

2

12

=

1

48

(11)

Quiz 4.10

(A) (1) For random variables X and Y from Example 4.1, we observe that P

Y

(1) =

0.09 and P

X

(0) = 0.01. However,

P

X,Y

(0, 1) = 0 = P

X

(0) P

Y

(1) (1)

Since we have found a pair x, y such that P

X,Y

(x, y) = P

X

(x)P

Y

(y), we can

conclude that X and Y are dependent. Note that whenever P

X,Y

(x, y) = 0,

independence requires that either P

X

(x) = 0 or P

Y

(y) = 0.

(2) For random variables Q and G from Quiz 4.2, it is not obvious whether they

are independent. Unlike X and Y in part (a), there are no obvious pairs q, g

that fail the independence requirement. In this case, we calculate the marginal

PMFs from the table of the joint PMF P

Q,G

(q, g) in Quiz 4.2.

P

Q,G

(q, g) g = 0 g = 1 g = 2 g = 3 P

Q

(q)

q = 0 0.06 0.18 0.24 0.12 0.60

q = 1 0.04 0.12 0.16 0.08 0.40

P

G

(g) 0.10 0.30 0.40 0.20

Careful study of the table will verify that P

Q,G

(q, g) = P

Q

(q)P

G

(g) for every

pair q, g. Hence Q and G are independent.

(B) (1) Since X

1

and X

2

are independent,

f

X

1

,X

2

(x

1

, x

2

) = f

X

1

(x

1

) f

X

2

(x

2

) (2)

=

¸

(1 − x

1

/2)(1 − x

2

/2) 0 ≤ x

1

≤ 2, 0 ≤ x

2

≤ 2

0 otherwise

(3)

30

(2) Let F

X

(x) denote the CDF of both X

1

and X

2

. The CDF of Z = max(X

1

, X

2

)

is found by observing that Z ≤ z iff X

1

≤ z and X

2

≤ z. That is,

P [Z ≤ z] = P [X

1

≤ z, X

2

≤ z] (4)

= P [X

1

≤ z] P [X

2

≤ z] = [F

X

(z)]

2

(5)

To complete the problem, we need to ﬁnd the CDF of each X

i

. From the PDF

f

X

(x), the CDF is

F

X

(x) =

x

−∞

f

X

(y) dy =

⎧

⎨

⎩

0 x < 0

x − x

2

/4 0 ≤ x ≤ 2

1 x > 2

(6)

Thus for 0 ≤ z ≤ 2,

F

Z

(z) = (z − z

2

/4)

2

(7)

The complete expression for the CDF of Z is

F

Z

(z) =

⎧

⎨

⎩

0 z < 0

(z − z

2

/4)

2

0 ≤ z ≤ 2

1 z > 1

(8)

Quiz 4.11

This problem just requires identifying the various terms in Deﬁnition 4.17 and Theo-

rem 4.29. Speciﬁcally, from the problem statement, we know that ρ = 1/2,

µ

1

= µ

X

= 0, µ

2

= µ

Y

= 0, (1)

and that

σ

1

= σ

X

= 1, σ

2

= σ

Y

= 1. (2)

(1) Applying these facts to Deﬁnition 4.17, we have

f

X,Y

(x, y) =

1

√

3π

2

e

−2(x

2

−xy+y

2

)/3

. (3)

(2) By Theorem 4.30, the conditional expected value and standard deviation of X given

Y = y are

E [X|Y = y] = y/2 ˜ σ

X

= σ

2

1

(1 −ρ

2

) =

3/4. (4)

When Y = y = 2, we see that E[X|Y = 2] = 1 and Var[X|Y = 2] = 3/4. The

conditional PDF of X given Y = 2 is simply the Gaussian PDF

f

X|Y

(x|2) =

1

√

3π/2

e

−2(x−1)

2

/3

. (5)

31

Quiz 4.12

One straightforward method is to follow the approach of Example 4.28. Instead, we use

an alternate approach. First we observe that X has the discrete uniform (1, 4) PMF. Also,

given X = x, Y has a discrete uniform (1, x) PMF. That is,

P

X

(x) =

¸

1/4 x = 1, 2, 3, 4,

0 otherwise,

P

Y|X

(y|x) =

¸

1/x y = 1, . . . , x

0 otherwise

(1)

Given X = x, and an independent uniform (0, 1) random variable U, we can generate a

sample value of Y with a discrete uniform (1, x) PMF via Y = xU. This observation

prompts the following program:

function xy=dtrianglerv(m)

sx=[1;2;3;4];

px=0.25*ones(4,1);

x=finiterv(sx,px,m);

y=ceil(x.*rand(m,1));

xy=[x’;y’];

32

Quiz Solutions – Chapter 5

Quiz 5.1

We ﬁnd P[C] by integrating the joint PDF over the region of interest. Speciﬁcally,

P [C] =

1/2

0

dy

2

y

2

0

dy

1

1/2

0

dy

4

y

4

0

4dy

3

(1)

= 4

1/2

0

y

2

dy

2

1/2

0

y

4

dy

4

= 1/4. (2)

Quiz 5.2

By deﬁnition of A, Y

1

= X

1

, Y

2

= X

2

−X

1

and Y

3

= X

3

−X

2

. Since 0 < X

1

< X

2

<

X

3

, each Y

i

must be a strictly positive integer. Thus, for y

1

, y

2

, y

3

∈ {1, 2, . . .},

P

Y

(y) = P [Y

1

= y

1

, Y

2

= y

2

, Y

3

= y

3

] (1)

= P [X

1

= y

1

, X

2

− X

1

= y

2

, X

3

− X

2

= y

3

] (2)

= P [X

1

= y

1

, X

2

= y

2

+ y

1

, X

3

= y

3

+ y

2

+ y

1

] (3)

= (1 − p)

3

p

y

1

+y

2

+y

3

(4)

By deﬁning the vector a =

¸

1 1 1

¸

**, the complete expression for the joint PMF of Y is
**

P

Y

(y) =

¸

(1 − p) p

a

y

y

1

, y

2

, y

3

∈ {1, 2, . . .}

0 otherwise

(5)

Quiz 5.3

First we note that each marginal PDF is nonzero only if any subset of the x

i

obeys the

ordering contraints 0 ≤ x

1

≤ x

2

≤ x

3

≤ 1. Within these constraints, we have

f

X

1

,X

2

(x

1

, x

2

) =

∞

−∞

f

X

(x) dx

3

=

1

x

2

6 dx

3

= 6(1 − x

2

), (1)

f

X

2

,X

3

(x

2

, x

3

) =

∞

−∞

f

X

(x) dx

1

=

x

2

0

6 dx

1

= 6x

2

, (2)

f

X

1

,X

3

(x

1

, x

3

) =

∞

−∞

f

X

(x) dx

2

=

x

3

x

1

6 dx

2

= 6(x

3

− x

1

). (3)

In particular, we must keep in mind that f

X

1

,X

2

(x

1

, x

2

) = 0 unless 0 ≤ x

1

≤ x

2

≤ 1,

f

X

2

,X

3

(x

2

, x

3

) = 0 unless 0 ≤ x

2

≤ x

3

≤ 1, and that f

X

1

,X

3

(x

1

, x

3

) = 0 unless 0 ≤ x

1

≤

33

x

3

≤ 1. The complete expressions are

f

X

1

,X

2

(x

1

, x

2

) =

¸

6(1 − x

2

) 0 ≤ x

1

≤ x

2

≤ 1

0 otherwise

(4)

f

X

2

,X

3

(x

2

, x

3

) =

¸

6x

2

0 ≤ x

2

≤ x

3

≤ 1

0 otherwise

(5)

f

X

1

,X

3

(x

1

, x

3

) =

¸

6(x

3

− x

1

) 0 ≤ x

1

≤ x

3

≤ 1

0 otherwise

(6)

Now we can ﬁnd the marginal PDFs. When 0 ≤ x

i

≤ 1 for each x

i

,

f

X

1

(x

1

) =

∞

−∞

f

X

1

,X

2

(x

1

, x

2

) dx

2

=

1

x

1

6(1 − x

2

) dx

2

= 3(1 − x

1

)

2

(7)

f

X

2

(x

2

) =

∞

−∞

f

X

2

,X

3

(x

2

, x

3

) dx

3

=

1

x

2

6x

2

dx

3

= 6x

2

(1 − x

2

) (8)

f

X

3

(x

3

) =

∞

−∞

f

X

2

,X

3

(x

2

, x

3

) dx

2

=

x

3

0

6x

2

dx

2

= 3x

2

3

(9)

The complete expressions are

f

X

1

(x

1

) =

¸

3(1 − x

1

)

2

0 ≤ x

1

≤ 1

0 otherwise

(10)

f

X

2

(x

2

) =

¸

6x

2

(1 − x

2

) 0 ≤ x

2

≤ 1

0 otherwise

(11)

f

X

3

(x

3

) =

¸

3x

2

3

0 ≤ x

3

≤ 1

0 otherwise

(12)

Quiz 5.4

In the PDF f

Y

(y), the components have dependencies as a result of the ordering con-

straints Y

1

≤ Y

2

and Y

3

≤ Y

4

. We can separate these constraints by creating the vectors

V =

¸

Y

1

Y

2

¸

, W =

¸

Y

3

Y

4

¸

. (1)

The joint PDF of V and W is

f

V,W

(v, w) =

¸

4 0 ≤ v

1

≤ v

2

≤ 1, 0 ≤ w

1

≤ w

2

≤ 1

0 otherwise

(2)

34

We must verify that V and W are independent. For 0 ≤ v

1

≤ v

2

≤ 1,

f

V

(v) =

f

V,W

(v, w) dw

1

dw

2

(3)

=

1

0

1

w

1

4 dw

2

dw

1

(4)

=

1

0

4(1 −w

1

) dw

1

= 2 (5)

Similarly, for 0 ≤ w

1

≤ w

2

≤ 1,

f

W

(w) =

f

V,W

(v, w) dv

1

dv

2

(6)

=

1

0

1

v

1

4 dv

2

dv

1

= 2 (7)

It follows that V and W have PDFs

f

V

(v) =

¸

2 0 ≤ v

1

≤ v

2

≤ 1

0 otherwise

, f

W

(w) =

¸

2 0 ≤ w

1

≤ w

2

≤ 1

0 otherwise

(8)

It is easy to verify that f

V,W

(v, w) = f

V

(v) f

W

(w), conﬁrming that V and W are indepen-

dent vectors.

Quiz 5.5

(A) Referring to Theorem 1.19, each test is a subexperiment with three possible out-

comes: L, A and R. In ﬁve trials, the vector X =

¸

X

1

X

2

X

3

¸

indicating the

number of outcomes of each subexperiment has the multinomial PMF

P

X

(x) =

⎧

⎨

⎩

5

x

1

,x

2

,x

3

(0.3)

x

1

(0.6)

x

2

(0.1)

x

3

x

1

+ x

2

+ x

3

= 5;

x

1

, x

2

, x

3

∈ {0, 1, . . . , 5}

0 otherwise

(1)

We can ﬁnd the marginal PMF for each X

i

from the joint PMF P

X

(x); however it

is simpler to just start from ﬁrst principles and observe that X

1

is the number of

occurrences of L in ﬁve independent tests. If we view each test as a trial with success

probability P[L] = 0.3, we see that X

1

is a binomial (n, p) = (5, 0.3) random

variable. Similarly, X

2

is a binomial (5, 0.6) random variable and X

3

is a binomial

(5, 0.1) random variable. That is, for p

1

= 0.3, p

2

= 0.6 and p

3

= 0.1,

P

X

i

(x) =

¸

5

x

p

x

i

(1 − p

i

)

5−x

x = 0, 1, . . . , 5

0 otherwise

(2)

35

From the marginal PMFs, we see that X

1

, X

2

and X

3

are not independent. Hence, we

must use Theorem 5.6 to ﬁnd the PMF of W. In particular, since X

1

+ X

2

+ X

3

= 5

and since each X

i

is non-negative, P

W

(0) = P

W

(1) = 0. Furthermore,

P

W

(2) = P

X

(1, 2, 2) + P

X

(2, 1, 2) + P

X

(2, 2, 1) (3)

=

5![0.3(0.6)

2

(0.1)

2

+0.3

2

(0.6)(0.1)

2

+0.3

2

(0.6)

2

(0.1)]

2!2!1!

(4)

= 0.1458 (5)

In addition, for w = 3, w = 4, and w = 5, the event W = w occurs if and only if

one of the mutually exclusive events X

1

= w, X

2

= w, or X

3

= w occurs. Thus,

P

W

(3) = P

X

1

(3) + P

X

2

(3) + P

X

3

(3) = 0.486 (6)

P

W

(4) = P

X

1

(4) + P

X

2

(4) + P

X

3

(4) = 0.288 (7)

P

W

(5) = P

X

1

(5) + P

X

2

(5) + P

X

3

(5) = 0.0802 (8)

(B) Since each Y

i

= 2X

i

+4, we can apply Theorem 5.10 to write

f

Y

(y) =

1

2

3

f

X

y

1

−4

2

,

y

2

−4

2

,

y

3

−4

2

(9)

=

¸

(1/8)e

−(y

3

−4)/2

4 ≤ y

1

≤ y

2

≤ y

3

0 otherwise

(10)

Note that for other matrices A, the constraints on y resulting from the constraints

0 ≤ X

1

≤ X

2

≤ X

3

can be much more complicated.

Quiz 5.6

We start by ﬁnding the components E[X

i

] =

∞

−∞

x f

X

i

(x) dx of µ

X

. To do so, we use

the marginal PDFs f

X

i

(x) found in Quiz 5.3:

E [X

1

] =

1

0

3x(1 − x)

2

dx = 1/4, (1)

E [X

2

] =

1

0

6x

2

(1 − x) dx = 1/2, (2)

E [X

3

] =

1

0

3x

3

dx = 3/4. (3)

To ﬁnd the correlation matrix R

X

, we need to ﬁnd E[X

i

X

j

] for all i and j . We start with

36

the second moments:

E

¸

X

2

1

¸

=

1

0

3x

2

(1 − x)

2

dx = 1/10. (4)

E

¸

X

2

2

¸

=

1

0

6x

3

(1 − x) dx = 3/10. (5)

E

¸

X

2

3

¸

=

1

0

3x

4

dx = 3/5. (6)

Using marginal PDFs from Quiz 5.3, the cross terms are

E [X

1

X

2

] =

∞

−∞

∞

−∞

x

1

x

2

f

X

1

,X

2

(x

1

, x

2

) , dx

1

dx

2

(7)

=

1

0

1

x

1

6x

1

x

2

(1 − x

2

) dx

2

dx

1

(8)

=

1

0

[x

1

−3x

3

1

+2x

4

1

] dx

1

= 3/20. (9)

E [X

2

X

3

] =

1

0

1

x

2

6x

2

2

x

3

dx

3

dx

2

(10)

=

1

0

[3x

2

2

−3x

4

2

] dx

2

= 2/5 (11)

E [X

1

X

3

] =

1

0

1

x

1

6x

1

x

3

(x

3

− x

1

) dx

3

dx

1

. (12)

=

1

0

(2x

1

x

3

3

−3x

2

1

x

2

3

)

x

3

=1

x

3

=x

1

dx

1

(13)

=

1

0

[2x

1

−3x

2

1

+ x

4

1

] dx

1

= 1/5. (14)

Summarizing the results, X has correlation matrix

R

X

=

⎡

⎣

1/10 3/20 1/5

3/20 3/10 2/5

1/5 2/5 3/5

⎤

⎦

. (15)

Vector X has covariance matrix

C

X

= R

X

− E [X] E [X]

(16)

=

⎡

⎣

1/10 3/20 1/5

3/20 3/10 2/5

1/5 2/5 3/5

⎤

⎦

−

⎡

⎣

1/4

1/2

3/4

⎤

⎦

¸

1/4 1/2 3/4

¸

(17)

=

⎡

⎣

1/10 3/20 1/5

3/20 3/10 2/5

1/5 2/5 3/5

⎤

⎦

−

⎡

⎣

1/16 1/8 3/16

1/8 1/4 3/8

3/16 3/8 9/16

⎤

⎦

=

1

80

⎡

⎣

3 2 1

2 4 2

1 2 3

⎤

⎦

. (18)

37

This problemshows that even for fairly simple joint PDFs, computing the covariance matrix

by calculus can be a time consuming task.

Quiz 5.7

We observe that X = AZ +b where

A =

¸

2 1

1 −1

¸

, b =

¸

2

0

¸

. (1)

It follows from Theorem 5.18 that µ

X

= b and that

C

X

= AA

=

¸

2 1

1 −1

¸ ¸

2 1

1 −1

¸

=

¸

5 1

1 2

¸

. (2)

Quiz 5.8

First, we observe that Y = AT where A =

¸

1/31 1/31 · · · 1/31

¸

. Since T is a

Gaussian random vector, Theorem 5.16 tells us that Y is a 1 dimensional Gaussian vector,

i.e., just a Gaussian random variable. The expected value of Y is µ

Y

= µ

T

= 80. The

covariance matrix of Y is 1 × 1 and is just equal to Var[Y]. Thus, by Theorem 5.16,

Var[Y] = AC

T

A

.

function p=julytemps(T);

[D1 D2]=ndgrid((1:31),(1:31));

CT=36./(1+abs(D1-D2));

A=ones(31,1)/31.0;

CY=(A’)*CT*A;

p=phi((T-80)/sqrt(CY));

In julytemps.m, the ﬁrst two lines gen-

erate the 31 ×31 covariance matrix CT, or

C

T

. Next we calculate Var[Y]. The ﬁnal

step is to use the (·) function to calculate

P[Y < T].

Here is the output of julytemps.m:

>> julytemps([70 75 80 85 90 95])

ans =

0.0000 0.0221 0.5000 0.9779 1.0000 1.0000

Note that P[T ≤ 70] is not actually zero and that P[T ≤ 90] is not actually 1.0000. Its

just that the MATLAB’s short format output, invoked with the command format short,

rounds off those probabilities. Here is the long format output:

>> format long

>> julytemps([70 75 80 85 90 95])

ans =

Columns 1 through 4

0.00002844263128 0.02207383067604 0.50000000000000 0.97792616932396

Columns 5 through 6

0.99997155736872 0.99999999922010

38

The ndgrid function is a useful to way calculate many covariance matrices. However, in

this problem, C

X

has a special structure; the i, j th element is

C

T

(i, j ) = c

|i −j |

=

36

1 +|i − j |

. (1)

If we write out the elements of the covariance matrix, we see that

C

T

=

⎡

⎢

⎢

⎢

⎣

c

0

c

1

· · · c

30

c

1

c

0

.

.

.

.

.

.

.

.

.

.

.

.

.

.

. c

1

c

30

· · · c

1

c

0

⎤

⎥

⎥

⎥

⎦

. (2)

This covariance matrix is known as a symmetric Toeplitz matrix. We will see in Chap-

ters 9 and 11 that Toeplitz covariance matrices are quite common. In fact, MATLAB has a

toeplitz function for generating them. The function julytemps2 use the toeplitz

to generate the correlation matrix C

T

.

function p=julytemps2(T);

c=36./(1+abs(0:30));

CT=toeplitz(c);

A=ones(31,1)/31.0;

CY=(A’)*CT*A;

p=phi((T-80)/sqrt(CY));

39

Quiz Solutions – Chapter 6

Quiz 6.1

Let K

1

, . . . , K

n

denote a sequence of iid random variables each with PMF

P

K

(k) =

¸

1/4 k = 1, . . . , 4

0 otherwise

(1)

We can write W

n

in the form of W

n

= K

1

+ · · · + K

n

. First, we note that the ﬁrst two

moments of K

i

are

E [K

i

] = (1 +2 +3 +4)/4 = 2.5 (2)

E

¸

K

2

i

¸

= (1

2

+2

2

+3

2

+4

2

)/4 = 7.5 (3)

Thus the variance of K

i

is

Var[K

i

] = E

¸

K

2

i

¸

−(E [K

i

])

2

= 7.5 −(2.5)

2

= 1.25 (4)

Since E[K

i

] = 2.5, the expected value of W

n

is

E [W

n

] = E [K

1

] +· · · + E [K

n

] = nE [K

i

] = 2.5n (5)

Since the rolls are independent, the random variables K

1

, . . . , K

n

are independent. Hence,

by Theorem 6.3, the variance of the sum equals the sum of the variances. That is,

Var[W

n

] = Var[K

1

] +· · · +Var[K

n

] = 1.25n (6)

Quiz 6.2

Random variables X and Y have PDFs

f

X

(x) =

¸

3e

−3x

x ≥ 0

0 otherwise

f

Y

(y) =

¸

2e

−2y

y ≥ 0

0 otherwise

(1)

Since X and Y are nonnegative, W = X +Y is nonnegative. By Theorem 6.5, the PDF of

W = X +Y is

f

W

(w) =

∞

−∞

f

X

(w − y) f

Y

(y) dy = 6

w

0

e

−3(w−y)

e

−2y

dy (2)

Fortunately, this integral is easy to evaluate. For w > 0,

f

W

(w) = e

−3w

e

y

w

0

= 6

e

−2w

−e

−3w

(3)

Since f

W

(w) = 0 for w < 0, a conmplete expression for the PDF of W is

f

W

(w) =

¸

6e

−2w

1 −e

−w

w ≥ 0,

0 otherwise.

(4)

40

Quiz 6.3

The MGF of K is

φ

K

(s) = E

¸

e

s K

¸

==

4

¸

k=0

(0.2)e

sk

= 0.2

1 +e

s

+e

2s

+e

3s

+e

4s

(1)

We ﬁnd the moments by taking derivatives. The ﬁrst derivative of φ

K

(s) is

dφ

K

(s)

ds

= 0.2(e

s

+2e

2s

+3e

3s

+4e

4s

) (2)

Evaluating the derivative at s = 0 yields

E [K] =

dφ

K

(s)

ds

s=0

= 0.2(1 +2 +3 +4) = 2 (3)

To ﬁnd higher-order moments, we continue to take derivatives:

E

¸

K

2

¸

=

d

2

φ

K

(s)

ds

2

s=0

= 0.2(e

s

+4e

2s

+9e

3s

+16e

4s

)

s=0

= 6 (4)

E

¸

K

3

¸

=

d

3

φ

K

(s)

ds

3

s=0

= 0.2(e

s

+8e

2s

+27e

3s

+64e

4s

)

s=0

= 20 (5)

E

¸

K

4

¸

=

d

4

φ

K

(s)

ds

4

s=0

= 0.2(e

s

+16e

2s

+81e

3s

+256e

4s

)

s=0

= 70.8 (6)

(7)

Quiz 6.4

(A) Each K

i

has MGF

φ

K

(s) = E

¸

e

s K

i

¸

=

e

s

+e

2s

+· · · +e

ns

n

=

e

s

(1 −e

ns

)

n(1 −e

s

)

(1)

Since the sequence of K

i

is independent, Theorem 6.8 says the MGF of J is

φ

J

(s) = (φ

K

(s))

m

=

e

ms

(1 −e

ns

)

m

n

m

(1 −e

s

)

m

(2)

(B) Since the set of α

j

X

j

are independent Gaussian random variables, Theorem 6.10

says that W is a Gaussian random variable. Thus to ﬁnd the PDF of W, we need

only ﬁnd the expected value and variance. Since the expectation of the sum equals

the sum of the expectations:

E [W] = αE [X

1

] +α

2

E [X

2

] +· · · +α

n

E [X

n

] = 0 (3)

41

Since the α

j

X

j

are independent, the variance of the sum equals the sum of the vari-

ances:

Var[W] = α

2

Var[X

1

] +α

4

Var[X

2

] +· · · +α

2n

Var[X

n

] (4)

= α

2

+2(α

2

)

2

+3(α

2

)

3

+· · · +n(α

2

)

n

(5)

Deﬁning q = α

2

, we can use Math Fact B.6 to write

Var[W] =

α

2

−α

2n+2

[1 +n(1 −α

2

)]

(1 −α

2

)

2

(6)

With E[W] = 0 and σ

2

W

= Var[W], we can write the PDF of W as

f

W

(w) =

1

2πσ

2

W

e

−w

2

/2σ

2

W

(7)

Quiz 6.5

(1) From Table 6.1, each X

i

has MGF φ

X

(s) and random variable N has MGF φ

N

(s)

where

φ

X

(s) =

1

1 −s

, φ

N

(s) =

1

5

e

s

1 −

4

5

e

s

. (1)

From Theorem 6.12, R has MGF

φ

R

(s) = φ

N

(ln φ

X

(s)) =

1

5

φ

X

(s)

1 −

4

5

φ

X

(s)

(2)

Substituting the expression for φ

X

(s) yields

φ

R

(s) =

1

5

1

5

−s

. (3)

(2) From Table 6.1, we see that R has the MGF of an exponential (1/5) random variable.

The corresponding PDF is

f

R

(r) =

¸

(1/5)e

−r/5

r ≥ 0

0 otherwise

(4)

This quiz is an example of the general result that a geometric sum of exponential

random variables is an exponential random variable.

42

Quiz 6.6

(1) The expected access time is

E [X] =

∞

−∞

x f

X

(x) dx =

12

0

x

12

dx = 6 msec (1)

(2) The second moment of the access time is

E

¸

X

2

¸

=

∞

−∞

x

2

f

X

(x) dx =

12

0

x

2

12

dx = 48 (2)

The variance of the access time is Var[X] = E[X

2

] −(E[X])

2

= 48 −36 = 12.

(3) Using X

i

to denote the access time of block i , we can write

A = X

1

+ X

2

+· · · + X

12

(3)

Since the expectation of the sum equals the sum of the expectations,

E [A] = E [X

1

] +· · · + E [X

12

] = 12E [X] = 72 msec (4)

(4) Since the X

i

are independent,

Var[A] = Var[X

1

] +· · · +Var[X

12

] = 12 Var[X] = 144 (5)

Hence, the standard deviation of A is σ

A

= 12

(5) To use the central limit theorem, we write

P [A > 75] = 1 − P [A ≤ 75] (6)

= 1 − P

¸

A − E [A]

σ

A

≤

75 − E [A]

σ

A

¸

(7)

≈ 1 −

75 −72

12

(8)

= 1 −0.5987 = 0.4013 (9)

Note that we used Table 3.1 to look up (0.25).

(6) Once again, we use the central limit theorem and Table 3.1 to estimate

P [A < 48] = P

¸

A − E [A]

σ

A

<

48 − E [A]

σ

A

¸

(10)

≈

48 −72

12

(11)

= 1 −(2) = 1 −0.9773 = 0.0227 (12)

43

Quiz 6.7

Random variable K

n

has a binomial distribution for n trials and success probability

P[V] = 3/4.

(1) The expected number of voice calls out of 48 calls is E[K

48

] = 48P[V] = 36.

(2) The variance of K

48

is

Var[K

48

] = 48P [V] (1 − P [V]) = 48(3/4)(1/4) = 9 (1)

Thus K

48

has standard deviation σ

K

48

= 3.

(3) Using the ordinary central limit theorem and Table 3.1 yields

P [30 ≤ K

48

≤ 42] ≈

42 −36

3

−

30 −36

3

= (2) −(−2) (2)

Recalling that (−x) = 1 −(x), we have

P [30 ≤ K

48

≤ 42] ≈ 2(2) −1 = 0.9545 (3)

(4) Since K

48

is a discrete random variable, we can use the De Moivre-Laplace approx-

imation to estimate

P [30 ≤ K

48

≤ 42] ≈

42 +0.5 −36

3

−

30 −0.5 −36

3

(4)

= 2(2.16666) −1 = 0.9687 (5)

Quiz 6.8

The train interarrival times X

1

, X

2

, X

3

are iid exponential (λ) random variables. The

arrival time of the third train is

W = X

1

+ X

2

+ X

3

. (1)

In Theorem 6.11, we found that the sum of three iid exponential (λ) random variables is an

Erlang (n = 3, λ) random variable. From Appendix A, we ﬁnd that W has expected value

and variance

E [W] = 3/λ = 6 Var[W] = 3/λ

2

= 12 (2)

(1) By the Central Limit Theorem,

P [W > 20] = P

¸

W −6

√

12

>

20 −6

√

12

¸

≈ Q(7/

√

3) = 2.66 ×10

−5

(3)

44

(2) To use the Chernoff bound, we note that the MGF of W is

φ

W

(s) =

λ

λ −s

3

=

1

(1 −2s)

3

(4)

The Chernoff bound states that

P [W > 20] ≤ min

s≥0

e

−20s

φ

X

(s) = min

s≥0

e

−20s

(1 −2s)

3

(5)

To minimize h(s) = e

−20s

/(1 −2s)

3

, we set the derivative of h(s) to zero:

dh(s)

ds

=

−20(1 −2s)

3

e

−20s

+6e

−20s

(1 −2s)

2

(1 −2s)

6

= 0 (6)

This implies 20(1 − 2s) = 6 or s = 7/20. Applying s = 7/20 into the Chernoff

bound yields

P [W > 20] ≤

e

−20s

(1 −2s)

3

s=7/20

= (10/3)

3

e

−7

= 0.0338 (7)

(3) Theorem 3.11 says that for any w > 0, the CDF of the Erlang (λ, 3) random variable

W satisﬁes

F

W

(w) = 1 −

2

¸

k=0

(λw)

k

e

−λw

k!

(8)

Equivalently, for λ = 1/2 and w = 20,

P [W > 20] = 1 − F

W

(20) (9)

= e

−10

1 +

10

1!

+

10

2

2!

= 61e

−10

= 0.0028 (10)

Although the Chernoff bound is relatively weak in that it overestimates the proba-

bility by roughly a factor of 12, it is a valid bound. By contrast, the Central Limit

Theorem approximation grossly underestimates the true probability.

Quiz 6.9

One solution to this problem is to follow the approach of Example 6.19:

%unifbinom100.m

sx=0:100;sy=0:100;

px=binomialpmf(100,0.5,sx); py=duniformpmf(0,100,sy);

[SX,SY]=ndgrid(sx,sy); [PX,PY]=ndgrid(px,py);

SW=SX+SY; PW=PX.*PY;

sw=unique(SW); pw=finitepmf(SW,PW,sw);

pmfplot(sw,pw,’\itw’,’\itP_W(w)’);

A graph of the PMF P

W

(w) appears in Figure 2 With some thought, it should be apparent

that the finitepmf function is implementing the convolution of the two PMFs.

45

0 20 40 60 80 100 120 140 160 180 200

0

0.002

0.004

0.006

0.008

0.01

w

P

W

(

w

)

Figure 2: From Quiz 6.9, the PMF P

W

(w) of the independent sum of a binomial (100, 0.5)

random variable and a discrete uniform (0, 100) random variable.

46

Quiz Solutions – Chapter 7

Quiz 7.1

An exponential random variable with expected value 1 also has variance 1. By Theo-

rem 7.1, M

n

(X) has variance Var[M

n

(X)] = 1/n. Hence, we need n = 100 samples.

Quiz 7.2

The arrival time of the third elevator is W = X

1

+ X

2

+ X

3

. Since each X

i

is uniform

(0, 30),

E [X

i

] = 15, Var [X

i

] =

(30 −0)

2

12

= 75. (1)

Thus E[W] = 3E[X

i

] = 45, and Var[W] = 3 Var[X

i

] = 225.

(1) By the Markov inequality,

P [W > 75] ≤

E [W]

75

=

45

75

=

3

5

(2)

(2) By the Chebyshev inequality,

P [W > 75] = P [W − E [W] > 30] (3)

≤ P [|W − E [W]| > 30] ≤

Var [W]

30

2

=

225

900

=

1

4

(4)

Quiz 7.3

Deﬁne the random variable W = (X − µ

X

)

2

. Observe that V

100

(X) = M

100

(W). By

Theorem 7.6, the mean square error is

E

¸

(M

100

(W) −µ

W

)

2

¸

=

Var[W]

100

(1)

Observe that µ

X

= 0 so that W = X

2

. Thus,

µ

W

= E

¸

X

2

¸

=

1

−1

x

2

f

X

(x) dx = 1/3 (2)

E

¸

W

2

¸

= E

¸

X

4

¸

=

1

−1

x

4

f

X

(x) dx = 1/5 (3)

Therefore Var[W] = E[W

2

] − µ

2

W

= 1/5 − (1/3)

2

= 4/45 and the mean square error is

4/4500 = 0.000889.

47

Quiz 7.4

Assuming the number n of samples is large, we can use a Gaussian approximation for

M

n

(X). SinceE[X] = p and Var[X] = p(1 − p), we apply Theorem 7.13 which says that

the interval estimate

M

n

(X) −c ≤ p ≤ M

n

(X) +c (1)

has conﬁdence coefﬁcient 1 −α where

α = 2 −2

c

√

n

p(1 − p)

. (2)

We must ensure for every value of p that 1 − α ≥ 0.9 or α ≤ 0.1. Equivalently, we must

have

c

√

n

p(1 − p)

≥ 0.95 (3)

for every value of p. Since (x) is an increasing function of x, we must satisfy c

√

n ≥

1.65p(1 − p). Since p(1 − p) ≤ 1/4 for all p, we require that

c ≥

1.65

4

√

n

=

0.41

√

n

. (4)

The 0.9 conﬁdence interval estimate of p is

M

n

(X) −

0.41

√

n

≤ p ≤ M

n

(X) +

0.41

√

n

. (5)

For the 0.99 conﬁdence interval, we have α ≤ 0.01, implying (c

√

n/( p(1−p))) ≥ 0.995.

This implies c

√

n ≥ 2.58p(1 − p). Since p(1 − p) ≤ 1/4 for all p, we require that

c ≥ (0.25)(2.58)/

√

n. In this case, the 0.99 conﬁdence interval estimate is

M

n

(X) −

0.645

√

n

≤ p ≤ M

n

(X) +

0.645

√

n

. (6)

Note that if M

100

(X) = 0.4, then the 0.99 conﬁdence interval estimate is

0.3355 ≤ p ≤ 0.4645. (7)

The interval is wide because the 0.99 conﬁdence is high.

Quiz 7.5

Following the approach of bernoullitraces.m, we generate m = 1000 sample

paths, each sample path having n = 100 Bernoulli traces. at time k, OK(k) counts the

fraction of sample paths that have sample mean within one standard error of p. The pro-

gram bernoullisample.m generates graphs the number of traces within one standard

error as a function of the time, i.e. the number of trials in each trace.

48

function OK=bernoullisample(n,m,p);

x=reshape(bernoullirv(p,m*n),n,m);

nn=(1:n)’*ones(1,m);

MN=cumsum(x)./nn;

stderr=sqrt(p*(1-p))./sqrt((1:n)’);

stderrmat=stderr*ones(1,m);

OK=sum(abs(MN-p)<stderrmat,2)/m;

plot(1:n,OK,’-s’);

The following graph was generated by bernoullisample(100,5000,0.5):

0 10 20 30 40 50 60 70 80 90 100

0.4

0.5

0.6

0.7

0.8

0.9

1

As we would expect, as m gets large, the fraction of traces within one standard error ap-

proaches 2(1) −1 ≈ 0.68. The unusual sawtooth pattern, though perhaps unexpected, is

examined in Problem 7.5.2.

49

Quiz Solutions – Chapter 8

Quiz 8.1

From the problem statement, each X

i

has PDF and CDF

f

X

i

(x) =

¸

e

−x

x ≥ 0

0 otherwise

F

X

i

(x) =

¸

0 x < 0

1 −e

−x

x ≥ 0

(1)

Hence, the CDF of the maximum of X

1

, . . . , X

15

obeys

F

X

(x) = P [X ≤ x] = P [X

1

≤ x, X

2

≤ x, · · · , X

15

≤ x] = [P [X

i

≤ x]]

15

. (2)

This implies that for x ≥ 0,

F

X

(x) =

¸

F

X

i

(x)

¸

15

=

¸

1 −e

−x

¸

15

(3)

To design a signiﬁcance test, we must choose a rejection region for X. A reasonable choice

is to reject the hypothesis if X is too small. That is, let R = {X ≤ r}. For a signiﬁcance

level of α = 0.01, we obtain

α = P [X ≤ r] = (1 −e

−r

)

15

= 0.01 (4)

It is straightforward to show that

r = −ln

¸

1 −(0.01)

1/15

¸

= 1.33 (5)

Hence, if we observe X < 1.33, then we reject the hypothesis.

Quiz 8.2

From the problem statement, the conditional PMFs of K are

P

K|H

0

(k) =

¸

10

4k

e

−10

4

k!

k = 0, 1, . . .

0 otherwise

(1)

P

K|H

1

(k) =

¸

10

6k

e

−10

6

k!

k = 0, 1, . . .

0 otherwise

(2)

Since the two hypotheses are equally likely, the MAP and ML tests are the same. From

Theorem 8.6, the ML hypothesis rule is

k ∈ A

0

if P

K|H

0

(k) ≥ P

K|H

1

(k) ; k ∈ A

1

otherwise. (3)

This rule simpliﬁes to

k ∈ A

0

if k ≤ k

∗

=

10

6

−10

4

ln 100

= 214, 975.7; k ∈ A

1

otherwise. (4)

Thus if we observe at least 214, 976 photons, then we accept hypothesis H

1

.

50

Quiz 8.3

For the QPSK system, a symbol error occurs when s

i

is transmitted but (X

1

, X

2

) ∈ A

j

for some j = i . For a QPSK system, it is easier to calculate the probability of a correct

decision. Given H

0

, the conditional probability of a correct decision is

P [C|H

0

] = P [X

1

> 0, X

2

> 0|H

0

] = P

¸

√

E/2 + N

1

> 0,

√

E/2 + N

2

> 0

¸

(1)

Because of the symmetry of the signals, P[C|H

0

] = P[C|H

i

] for all i . This implies the

probability of a correct decision is P[C] = P[C|H

0

]. Since N

1

and N

2

are iid Gaussian

(0, σ) random variables, we have

P [C] = P [C|H

0

] = P

¸

√

E/2 + N

1

> 0

¸

P

¸

√

E/2 + N

2

> 0

¸

(2)

=

P

¸

N

1

> −

√

E/2

¸

2

(3)

=

¸

1 −

−

√

E/2

σ

2

(4)

Since (−x) = 1 − (x), we have P[C] =

2

(

E/2σ

2

). Equivalently, the probability

of error is

P

ERR

= 1 − P [C] = 1 −

2

E

2σ

2

(5)

Quiz 8.4

To generate the ROC, the existing program sqdistor already calculates this miss

probability P

MISS

= P

01

and the false alarm probability P

FA

= P

10

. The modiﬁed pro-

gram, sqdistroc.m is essentially the same as sqdistor except the output is a ma-

trix FM whose columns are the false alarm and miss probabilities. Next, the program

sqdistrocplot.m calls sqdistroc three times to generate a plot that compares the

receiver performance for the three requested values of d. Here is the modiﬁed code:

function FM=sqdistroc(v,d,m,T)

%square law distortion recvr

%P(error) for m bits tested

%transmit v volts or -v volts,

%add N volts, N is Gauss(0,1)

%add d(v+N)ˆ2 distortion

%receive 1 if x>T, otherwise 0

%FM = [P(FA) P(MISS)]

x=(v+randn(m,1));

[XX,TT]=ndgrid(x,T(:));

P01=sum((XX+d*(XX.ˆ2)< TT),1)/m;

x= -v+randn(m,1);

[XX,TT]=ndgrid(x,T(:));

P10=sum((XX+d*(XX.ˆ2)>TT),1)/m;

FM=[P10(:) P01(:)];

function FM=sqdistrocplot(v,m,T);

FM1=sqdistroc(v,0.1,m,T);

FM2=sqdistroc(v,0.2,m,T);

FM5=sqdistroc(v,0.3,m,T);

FM=[FM1 FM2 FM5];

loglog(FM1(:,1),FM1(:,2),’-k’, ...

FM2(:,1),FM2(:,2),’--k’, ...

FM5(:,1),FM5(:,2),’:k’);

legend(’\it d=0.1’,’\it d=0.2’,...

’\it d=0.3’,3)

ylabel(’P_{MISS}’);

xlabel(’P_{FA}’);

51

To see the effect of d, the commands

T=-3:0.1:3; sqdistrocplot(3,100000,T);

generated the plot shown in Figure 3.

10

−5

10

−4

10

−3

10

−2

10

−1

10

0

10

−5

10

−4

10

−3

10

−2

10

−1

10

0

P

M

I

S

S

P

FA

d=0.1

d=0.2

d=0.3

T=-3:0.1:3; sqdistrocplot(3,100000,T);

Figure 3: The receiver operating curve for the communications system of Quiz 8.4 with

squared distortion.

52

Quiz Solutions – Chapter 9

Quiz 9.1

(1) First, we calculate the marginal PDF for 0 ≤ y ≤ 1:

f

Y

(y) =

y

0

2(y + x) dx = 2xy + x

2

x=y

x=0

= 3y

2

(1)

This implies the conditional PDF of X given Y is

f

X|Y

(x|y) =

f

X,Y

(x, y)

f

Y

(y)

=

¸

2

3y

+

2x

3y

2

0 ≤ x ≤ y

0 otherwise

(2)

(2) The minimum mean square error estimate of X given Y = y is

ˆ x

M

(y) = E [X|Y = y] =

y

0

2x

3y

+

2x

2

3y

2

dx = 5y/9 (3)

Thus the MMSE estimator of X given Y is

ˆ

X

M

(Y) = 5Y/9.

(3) To obtain the conditional PDF f

Y|X

(y|x), we need the marginal PDF f

X

(x). For

0 ≤ x ≤ 1,

f

X

(x) =

1

x

2(y + x) dy = y

2

+2xy

y=1

y=x

= 1 +2x −3x

2

(4)

(5)

For 0 ≤ x ≤ 1, the conditional PDF of Y given X is

f

Y|X

(y|x) =

¸

2(y+x)

1+2x−3x

2

x ≤ y ≤ 1

0 otherwise

(6)

(4) The MMSE estimate of Y given X = x is

ˆ y

M

(x) = E [Y|X = x] =

1

x

2y

2

+2xy

1 +2x −3x

2

dy (7)

=

2y

3

/3 + xy

2

1 +2x −3x

2

y=1

y=x

(8)

=

2 +3x −5x

3

3 +6x −9x

2

(9)

53

Quiz 9.2

(1) Since the expectation of the sum equals the sum of the expectations,

E [R] = E [T] + E [X] = 0 (1)

(2) Since T and X are independent, the variance of the sum R = T + X is

Var[R] = Var[T] +Var[X] = 9 +3 = 12 (2)

(3) Since T and R have expected values E[R] = E[T] = 0,

Cov [T, R] = E [T R] = E [T(T + X)] = E

¸

T

2

¸

+ E [T X] (3)

Since T and X are independent and have zero expected value, E[T X] = E[T]E[X] =

0 and E[T

2

] = Var[T]. Thus Cov[T, R] = Var[T] = 9.

(4) From Deﬁnition 4.8, the correlation coefﬁcient of T and R is

ρ

T,R

=

Cov [T, R]

√

Var[R] Var[T]

=

σ

T

σ

R

=

√

3/2 (4)

(5) From Theorem 9.4, the optimum linear estimate of T given R is

ˆ

T

L

(R) = ρ

T,R

σ

T

σ

R

(R − E [R]) + E [T] (5)

Since E[R] = E[T] = 0 and ρ

T,R

= σ

T

/σ

R

,

ˆ

T

L

(R) =

σ

2

T

σ

2

R

R =

σ

2

T

σ

2

T

+σ

2

X

R =

3

4

R (6)

Hence a

∗

= 3/4 and b

∗

= 0.

(6) By Theorem 9.4, the mean square error of the linear estimate is

e

∗

L

= Var[T](1 −ρ

2

T,R

) = 9(1 −3/4) = 9/4 (7)

Quiz 9.3

When R = r, the conditional PDF of X = Y −40−40 log

10

r is Gaussian with expected

value −40 −40 log

10

r and variance 64. The conditional PDF of X given R is

f

X|R

(x|r) =

1

√

128π

e

−(x+40+40 log

10

r)

2

/128

(1)

54

From the conditional PDF f

X|R

(x|r), we can use Deﬁnition 9.2 to write the ML estimate

of R given X = x as

ˆ r

ML

(x) = arg max

r≥0

f

X|R

(x|r) (2)

We observe that f

X|R

(x|r) is maximized when the exponent (x + 40 + 40 log

10

r)

2

is

minimized. This minimum occurs when the exponent is zero, yielding

log

10

r = −1 − x/40 (3)

or

ˆ r

ML

(x) = (0.1)10

−x/40

m (4)

If the result doesn’t look correct, note that a typical ﬁgure for the signal strength might be

x = −120 dB. This corresponds to a distance estimate of ˆ r

ML

(−120) = 100 m.

For the MAP estimate, we observe that the joint PDF of X and R is

f

X,R

(x, r) = f

X|R

(x|r) f

R

(r) =

1

10

6

√

32π

re

−(x+40+40 log

10

r)

2

/128

(5)

From Theorem 9.6, the MAP estimate of R given X = x is the value of r that maximizes

f

X,R

(x, r). That is,

ˆ r

MAP

(x) = arg max

0≤r≤1000

f

X,R

(x, r) (6)

Note that we have included the constraint r ≤ 1000 in the maximization to highlight the

fact that under our probability model, R ≤ 1000 m. Setting the derivative of f

X,R

(x, r)

with respect to r to zero yields

e

−(x+40+40 log

10

r)

2

/128

¸

1 −

80 log

10

e

128

(x +40 +40 log

10

r)

¸

= 0 (7)

Solving for r yields

r = 10

1

25 log

10

e

−1

10

−x/40

= (0.1236)10

−x/40

(8)

This is the MAP estimate of R given X = x as long as r ≤ 1000 m. When x ≤ −156.3 dB,

the above estimate will exceed 1000 m, which is not possible in our probability model.

Hence, the complete description of the MAP estimate is

ˆ r

MAP

(x) =

¸

1000 x < −156.3

(0.1236)10

−x/40

x ≥ −156.3

(9)

For example, if x = −120dB, then ˆ r

MAP

(−120) = 123.6 m. When the measured signal

strength is not too low, the MAP estimate is 23.6% larger than the ML estimate. This re-

ﬂects the fact that large values of R are a priori more probable than small values. However,

for very low signal strengths, the MAP estimate takes into account that the distance can

never exceed 1000 m.

55

Quiz 9.4

(1) From Theorem 9.4, the LMSE estimate of X

2

given Y

2

is

ˆ

X

2

(Y

2

) = a

∗

Y

2

+b

∗

where

a

∗

=

Cov [X

2

, Y

2

]

Var[Y

2

]

, b

∗

= µ

X

2

−a

∗

µ

Y

2

. (1)

Because E[X] = E[Y] = 0,

Cov [X

2

, Y

2

] = E [X

2

Y

2

] = E [X

2

(X

2

+ W

2

)] = E

¸

X

2

2

¸

= 1 (2)

Var[Y

2

] = Var[X

2

] +Var[W

2

] = E

¸

X

2

2

¸

+ E

¸

W

2

2

¸

= 1.1 (3)

It follows that a

∗

= 1/1.1. Because µ

X

2

= µ

Y

2

= 0, it follows that b

∗

= 0. Finally,

to compute the expected square error, we calculate the correlation coefﬁcient

ρ

X

2

,Y

2

=

Cov [X

2

, Y

2

]

σ

X

2

σ

Y

2

=

1

√

1.1

(4)

The expected square error is

e

∗

L

= Var[X

2

](1 −ρ

2

X

2

,Y

2

) = 1 −

1

1.1

=

1

11

= 0.0909 (5)

(2) Since Y = X + W and E[X] = E[W] = 0, it follows that E[Y] = 0. Thus we can

apply Theorem 9.7. Note that X and W have correlation matrices

R

X

=

¸

1 −0.9

−0.9 1

¸

, R

W

=

¸

0.1 0

0 0.1

¸

. (6)

In terms of Theorem 9.7, n = 2 and we wish to estimate X

2

given the observation

vector Y =

¸

Y

1

Y

2

¸

**. To apply Theorem 9.7, we need to ﬁnd R
**

Y

and R

YX

2

.

R

Y

= E

¸

YY

¸

= E

¸

(X +W)(X

+W

)

¸

(7)

= E

¸

XX

+XW

+WX

+WW

¸

. (8)

Because Xand Ware independent, E[XW

] = E[X]E[W

] = 0. Similarly, E[WX

] =

0. This implies

R

Y

= E

¸

XX

¸

+ E

¸

WW

¸

= R

X

+R

W

=

¸

1.1 −0.9

−0.9 1.1

¸

. (9)

In addition, we need to ﬁnd

R

YX

2

= E [YX

2

] =

¸

E [Y

1

X

2

]

E [Y

2

X

2

]

¸

=

¸

E [(X

1

+ W

1

)X

2

]

E [(X

2

+ W

2

)X

2

]

¸

. (10)

56

Since Xand Ware independent vectors, E[W

1

X

2

] = E[W

1

]E[X

2

] = 0 and E[W

2

X

2

] =

0. Thus

R

YX

2

=

¸

E[X

1

X

2

]

E

¸

X

2

2

¸

¸

=

¸

−0.9

1

¸

. (11)

By Theorem 9.7,

ˆ a = R

−1

Y

R

YX

2

=

¸

−0.225

0.725

¸

(12)

Therefore, the optimum linear estimator of X

2

given Y

1

and Y

2

is

ˆ

X

L

= ˆ a

Y = −0.225Y

1

+0.725Y

2

. (13)

The mean square error is

Var [X

2

] − ˆ a

R

YX

2

= Var [X] −a

1

r

Y

1

,X

2

−a

2

r

Y

2

,X

2

= 0.0725. (14)

Quiz 9.5

Since X and W have zero expected value, Y also has zero expected value. Thus, by

Theorem 9.7,

ˆ

X

L

(Y) = ˆ a

Y where ˆ a = R

−1

Y

R

YX

. Since X and W are independent,

E[WX] = 0 and E[XW

] = 0

. This implies

R

YX

= E [YX] = E [(1X +W)X] = 1E

¸

X

2

¸

= 1. (1)

By the same reasoning, the correlation matrix of Y is

R

Y

= E

¸

YY

¸

= E

¸

(1X +W)(1

X +W

)

¸

(2)

= 11

E

¸

X

2

¸

+1E

¸

XW

¸

+ E [WX] 1

+ E

¸

WW

¸

(3)

= 11

+R

W

(4)

Note that 11

**is a 20 ×20 matrix with every entry equal to 1. Thus,
**

ˆ a = R

−1

Y

R

YX

=

11

+R

W

−1

1 (5)

and the optimal linear estimator is

ˆ

X

L

(Y) = 1

11

+R

W

−1

Y (6)

The mean square error is

e

∗

L

= Var[X] − ˆ a

R

YX

= 1 −1

11

+R

W

−1

1 (7)

Now we note that R

W

has i, j th entry R

W

(i, j ) = c

|i −j |−1

. The question we must address

is what value c minimizes e

∗

L

. This problem is atypical in that one does not usually get

57

to choose the correlation structure of the noise. However, we will see that the answer is

somewhat instructive.

We note that the answer is not obviously apparent from Equation (7). In particular, we

observe that Var[W

i

] = R

W

(i, i ) = 1/c. Thus, when c is small, the noises W

i

have high

variance and we would expect our estimator to be poor. On the other hand, if c is large

W

i

and W

j

are highly correlated and the separate measurements of X are very dependent.

This would suggest that large values of c will also result in poor MSE. If this argument is

not clear, consider the extreme case in which every W

i

and W

j

have correlation coefﬁcient

ρ

i j

= 1. In this case, our 20 measurements will be all the same and one measurement is as

good as 20 measurements.

To ﬁnd the optimal value of c, we write a MATLAB function mquiz9(c) to calculate

the MSE for a given c and second function that ﬁnds plots the MSE for a range of values

of c.

function [mse,af]=mquiz9(c);

v1=ones(20,1);

RW=toeplitz(c.ˆ((0:19)-1));

RY=(v1*(v1’)) +RW;

af=(inv(RY))*v1;

mse=1-((v1’)*af);

function cmin=mquiz9minc(c);

msec=zeros(size(c));

for k=1:length(c),

[msec(k),af]=mquiz9(c(k));

end

plot(c,msec);

xlabel(’c’);ylabel(’e_Lˆ*’);

[msemin,optk]=min(msec);

cmin=c(optk);

Note in mquiz9 that v1 corresponds to the vector 1 of all ones. The following commands

ﬁnds the minimum c and also produces the following graph:

>> c=0.01:0.01:0.99;

>> mquiz9minc(c)

ans =

0.4500

0 0.5 1

0.2

0.4

0.6

0.8

1

c

e

L *

As we see in the graph, both small values and large values of c result in large MSE.

58

Quiz Solutions – Chapter 10

Quiz 10.1

There are many correct answers to this question. A correct answer speciﬁes enough

random variables to specify the sample path exactly. One choice for an alternate set of

random variables that would specify m(t, s) is

• m(0, s), the number of ongoing calls at the start of the experiment

• N, the number of new calls that arrive during the experiment

• X

1

, . . . , X

N

, the interarrival times of the N new arrivals

• H, the number of calls that hang up during the experiment

• D

1

, . . . , D

H

, the call completion times of the H calls that hang up

Quiz 10.2

(1) We obtain a continuous time, continuous valued process when we record the temper-

ature as a continuous waveform over time.

(2) If at every moment in time, we round the temperature to the nearest degree, then we

obtain a continuous time, discrete valued process.

(3) If we sample the process in part (a) every T seconds, then we obtain a discrete time,

continuous valued process.

(4) Rounding the samples in part (c) to the nearest integer degree yields a discrete time,

discrete valued process.

Quiz 10.3

(1) Each resistor has resistance R in ohms with uniform PDF

f

R

(r) =

¸

0.01 950 ≤ r ≤ 1050

0 otherwise

(1)

The probability that a test produces a 1% resistor is

p = P [990 ≤ R ≤ 1010] =

1010

990

(0.01) dr = 0.2 (2)

59

(2) In t seconds, exactly t resistors are tested. Each resistor is a 1% resistor with proba-

bility p, independent of any other resistor. Consequently, the number of 1% resistors

found has the binomial PMF

P

N(t )

(n) =

¸

t

n

p

n

(1 − p)

t −n

n = 0, 1, . . . , t

0 otherwise

(3)

(3) First we will ﬁnd the PMF of T

1

. This problem is easy if we view each resistor test

as an independent trial. A success occurs on a trial with probability p if we ﬁnd a

1% resistor. The ﬁrst 1% resistor is found at time T

1

= t if we observe failures on

trials 1, . . . , t − 1 followed by a success on trial t . Hence, just as in Example 2.11,

T

1

has the geometric PMF

P

T

1

(t ) =

¸

(1 − p)

t −1

p t = 1, 2, . . .

9 otherwise

(4)

Since p = 0.2, the probability the ﬁrst 1% resistor is found in exactly ﬁve seconds is

P

T

1

(5) = (0.8)

4

(0.2) = 0.08192.

(4) From Theorem 2.5, a geometric random variable with success probability p has ex-

pected value 1/p. In this problem, E[T

1

] = 1/p = 5.

(5) Note that once we ﬁnd the ﬁrst 1% resistor, the number of additional trials needed to

ﬁnd the second 1% resistor once again has a geometric PMF with expected value 1/p

since each independent trial is a success with probability p. That is, T

2

= T

1

+ T

where T

**is independent and identically distributed to T
**

1

. Thus

E [T

2

|T

1

= 10] = E [T

1

|T

1

= 10] + E

¸

T

|T

1

= 10

¸

(5)

= 10 + E

¸

T

¸

= 10 +5 = 15 (6)

Quiz 10.4

Since each X

i

is a N(0, 1) random variable, each X

i

has PDF

f

X(i )

(x) =

1

√

2π

e

−x

2

/2

(1)

By Theorem 10.1, the joint PDF of X =

¸

X

1

· · · X

n

¸

is

f

X

(x) = f

X(1),...,X(n)

(x

1

, . . . , x

n

) =

k

¸

i =1

f

X

(x

i

) =

1

(2π)

n/2

e

−(x

2

1

+···+x

2

n

)/2

(2)

60

Quiz 10.5

The ﬁrst and second hours are nonoverlapping intervals. Since one hour equals 3600

sec and the Poisson process has a rate of 10 packets/sec, the expected number of packets

in each hour is E[M

i

] = α = 36, 000. This implies M

1

and M

2

are independent Poisson

random variables each with PMF

P

M

i

(m) =

¸

α

m

e

−α

m!

m = 0, 1, 2, . . .

0 otherwise

(1)

Since M

1

and M

2

are independent, the joint PMF of M

1

and M

2

is

P

M

1

,M

2

(m

1

, m

2

) = P

M

1

(m

1

) P

M

2

(m

2

) =

⎧

⎪

⎪

⎨

⎪

⎪

⎩

α

m

1

+m

2

e

−2α

m

1

!m

2

!

m

1

= 0, 1, . . . ;

m

2

= 0, 1, . . . ,

0 otherwise.

(2)

Quiz 10.6

To answer whether N

**(t ) is a Poisson process, we look at the interarrival times. Let
**

X

1

, X

2

, . . . denote the interarrival times of the N(t ) process. Since we count only even-

numbered arrival for N

(t ), the time until the ﬁrst arrival of the N

(t ) is Y

1

= X

1

+ X

2

.

Since X

1

and X

2

are independent exponential (λ) random variables, Y

1

is an Erlang (n =

2, λ) random variable; see Theorem 6.11. Since Y

i

(t ), the i th interarrival time of the N

(t )

process, has the same PDF as Y

1

(t ), we can conclude that the interarrival times of N

(t )

are not exponential random variables. Thus N

**(t ) is not a Poisson process.
**

Quiz 10.7

First, we note that for t > s,

X(t ) − X(s) =

W(t ) − W(s)

√

α

(1)

Since W(t ) −W(s) is a Gaussian random variable, Theorem 3.13 states that W(t ) −W(s)

is Gaussian with expected value

E [X(t ) − X(s)] =

E [W(t ) − W(s)]

√

α

= 0 (2)

and variance

E

¸

(W(t ) − W(s))

2

¸

=

E

¸

(W(t ) − W(s))

2

¸

α

=

α(t −s)

α

(3)

Consider s

≤ s < t . Since s ≥ s

, W(t ) − W(s) is independent of W(s

). This implies

[W(t ) − W(s)]/

√

α is independent of W(s

)/

√

α for all s ≥ s

. That is, X(t ) − X(s) is

independent of X(s

) for all s ≥ s

**. Thus X(t ) is a Brownian motion process with variance
**

Var[X(t )] = t .

61

Quiz 10.8

First we ﬁnd the expected value

µ

Y

(t ) = µ

X

(t ) +µ

N

(t ) = µ

X

(t ). (1)

To ﬁnd the autocorrelation, we observe that since X(t ) and N(t ) are independent and since

N(t ) has zero expected value, E[X(t )N(t

)] = E[X(t )]E[N(t

)] = 0. Since R

Y

(t, τ) =

E[Y(t )Y(t +τ)], we have

R

Y

(t, τ) = E [(X(t ) + N(t )) (X(t +τ) + N(t +τ))] (2)

= E [X(t )X(t +τ)] + E [X(t )N(t +τ)]

+ E [X(t +τ)N(t )] + E [N(t )N(t +τ)] (3)

= R

X

(t, τ) + R

N

(t, τ). (4)

Quiz 10.9

From Deﬁnition 10.14, X

1

, X

2

, . . . is a stationary random sequence if for all sets of

time instants n

1

, . . . , n

m

and time offset k,

f

X

n

1

,...,X

n

m

(x

1

, . . . , x

m

) = f

X

n

1

+k

,...,X

n

m

+k

(x

1

, . . . , x

m

) (1)

Since the random sequence is iid,

f

X

n

1

,...,X

n

m

(x

1

, . . . , x

m

) = f

X

(x

1

) f

X

(x

2

) · · · f

X

(x

m

) (2)

Similarly, for time instants n

1

+k, . . . , n

m

+k,

f

X

n

1

+k

,...,X

n

m

+k

(x

1

, . . . , x

m

) = f

X

(x

1

) f

X

(x

2

) · · · f

X

(x

m

) (3)

We can conclude that the iid random sequence is stationary.

Quiz 10.10

We must check whether each function R(τ) meets the conditions of Theorem 10.12:

R(τ) ≥ 0 R(τ) = R(−τ) |R(τ)| ≤ R(0) (1)

(1) R

1

(τ) = e

−|τ|

meets all three conditions and thus is valid.

(2) R

2

(τ) = e

−τ

2

also is valid.

(3) R

3

(τ) = e

−τ

cos τ is not valid because

R

3

(−2π) = e

2π

cos 2π = e

2π

> 1 = R

3

(0) (2)

(4) R

4

(τ) = e

−τ

2

sin τ also cannot be an autocorrelation function because

R

4

(π/2) = e

−π/2

sin π/2 = e

−π/2

> 0 = R

4

(0) (3)

62

Quiz 10.11

(1) The autocorrelation of Y(t ) is

R

Y

(t, τ) = E [Y(t )Y(t +τ)] (1)

= E [X(−t )X(−t −τ)] (2)

= R

X

(−t −(−t −τ)) = R

X

(τ) (3)

Since E[Y(t )] = E[X(−t )] = µ

X

, we can conclude that Y(t ) is a wide sense

stationary process. In fact, we see that by viewing a process backwards in time, we

see the same second order statistics.

(2) Since X(t ) and Y(t ) are both wide sense stationary processes, we can check whether

they are jointly wide sense stationary by seeing if R

XY

(t, τ) is just a function of τ.

In this case,

R

XY

(t, τ) = E [X(t )Y(t +τ)] (4)

= E [X(t )X(−t −τ)] (5)

= R

X

(t −(−t −τ)) = R

X

(2t +τ) (6)

Since R

XY

(t, τ) depends on both t and τ, we conclude that X(t ) and Y(t ) are not

jointly wide sense stationary. To see why this is, suppose R

X

(τ) = e

−|τ|

so that

samples of X(t ) far apart in time have almost no correlation. In this case, as t gets

larger, Y(t ) = X(−t ) and X(t ) become less and less correlated.

Quiz 10.12

From the problem statement,

E [X(t )] = E [X(t +1)] = 0 (1)

E [X(t )X(t +1)] = 1/2 (2)

Var[X(t )] = Var[X(t +1)] = 1 (3)

The Gaussian random vector X =

¸

X(t ) X(t +1)

¸

**has covariance matrix and corre-
**

sponding inverse

C

X

=

¸

1 1/2

1/2 1

¸

C

−1

X

=

4

3

¸

1 −1/2

−1/2 1

¸

(4)

Since

x

C

−1

X

x =

¸

x

0

x

1

¸

4

3

¸

1 −1/2

−1/2 1

¸ ¸

x

0

x

1

¸

=

4

3

x

2

0

− x

0

x

+

x

2

1

(5)

the joint PDF of X(t ) and X(t +1) is the Gaussian vector PDF

f

X(t ),X(t +1)

(x

0

, x

1

) =

1

(2π)

n/2

[det (C

X

)]

1/2

exp

−

1

2

x

C

−1

X

x

(6)

=

1

√

3π

2

e

−

2

3

x

2

0

−x

0

x

1

+x

2

1

(7)

63

0 10 20 30 40 50 60 70 80 90 100

0

20

40

60

80

100

120

t

M

(

t

)

Figure 4: Sample path of 100 minutes of the blocking switch of Quiz 10.13.

Quiz 10.13

The simple structure of the switch simulation of Example 10.28 admits a deceptively

simple solution in terms of the vector of arrivals A and the vector of departures D. With the

introduction of call blocking. we cannot generate these vectors all at once. In particular,

when an arrival occurs at time t , we need to know that M(t ), the number of ongoing calls,

satisﬁes M(t ) < c = 120. Otherwise, when M(t ) = c, we must block the call. Call

blocking can be implemented by setting the service time of the call to zero so that the call

departs as soon as it arrives.

The blocking switch is an example of a discrete event system. The system evolves via

a sequence of discrete events, namely arrivals and departures, at discrete time instances. A

simulation of the system moves from one time instant to the next by maintaining a chrono-

logical schedule of future events (arrivals and departures) to be executed. The program

simply executes the event at the head of the schedule. The logic of such a simulation is

1. Start at time t = 0 with an empty system. Schedule the ﬁrst arrival to occur at S

1

, an

exponential (λ) random variable.

2. Examine the head-of-schedule event.

• When the head-of-schedule event is the kth arrival is at time t , check the state

M(t ).

– If M(t ) < c, admit the arrival, increase the system state n by 1, and sched-

ule a departure to occur at time t + S

n

, where S

k

is an exponential (λ)

random variable.

– If M(t ) = c, block the arrival, do not schedule a departure event.

• If the head of schedule event is a departure, reduce the system state n by 1.

3. Delete the head-of-schedule event and go to step 2.

After the head-of-schedule event is completed and any new events (departures in this sys-

tem) are scheduled, we know the system state cannot change until the next scheduled event.

64

Thus we know that M(t ) will stay the same until then. In our simulation, we use the vector

t as the set of time instances at which we inspect the system state. Thus for all times t(i)

between the current head-of-schedule event and the next, we set m(i) to the current switch

state.

The complete program is shown in Figure 5. In most programming languages, it is

common to implement the event schedule as a linked list where each item in the list has

a data structure indicating an event timestamp and the type of the event. In MATLAB, a

simple (but not elegant) way to do this is to have maintain two vectors: time is a list

of timestamps of scheduled events and event is a the list of event types. In this case,

event(i)=1 if the i th scheduled event is an arrival, or event(i)=-1 if the i th sched-

uled event is a departure.

When the program is passed a vector t, the output [m a b] is such that m(i) is the

number of ongoing calls at time t(i) while a and b are the number of admits and blocks.

The following instructions

t=0:0.1:5000;

[m,a,b]=simblockswitch(10,0.1,120,t);

plot(t,m);

generated a simulation lasting 5,000 minutes. A sample path of the ﬁrst 100 minutes of

that simulation is shown in Figure 4. The 5,000 minute full simulation produced a=49658

admitted calls and b=239 blocked calls. We can estimate the probability a call is blocked

as

ˆ

P

b

=

b

a +b

= 0.0048. (1)

In Chapter 12, we will learn that the exact blocking probability is given by Equation (12.93),

a result known as the “Erlang-B formula.” From the Erlang-B formula, we can calculate

that the exact blocking probability is P

b

= 0.0057. One reason our simulation underesti-

mates the blocking probability is that in a 5,000 minute simulation, roughly the ﬁrst 100

minutes are needed to load up the switch since the switch is idle when the simulation starts

at time t = 0. However, this says that roughly the ﬁrst two percent of the simulation time

was unusual. Thus this would account for only part of the disparity. The rest of the gap

between 0.0048 and 0.0057 is that a simulation that includes only 239 blocks is not all that

likely to give a very accurate result for the blocking probability.

Note that in Chapter 12, we will learn that the blocking switch is an example of an

M/M/c/c queue, a kind of Markov chain. Chapter 12 develops techniques for analyzing

and simulating systems described by Markov chains that are much simpler than the discrete

event simulation technique shown here. Nevertheless, for very complicated systems, the

discrete event simulation is widely-used and often very efﬁcient simulation method.

65

function [M,admits,blocks]=simblockswitch(lam,mu,c,t);

blocks=0; %total # blocks

admits=0; %total # admits

M=zeros(size(t));

n=0; % # in system

time=[ exponentialrv(lam,1) ];

event=[ 1 ]; %first event is an arrival

timenow=0;

tmax=max(t);

while (timenow<tmax)

M((timenow<=t)&(t<time(1)))=n;

timenow=time(1);

eventnow=event(1);

event(1)=[ ]; time(1)= [ ]; % clear current event

if (eventnow==1) % arrival

arrival=timenow+exponentialrv(lam,1); % next arrival

b4arrival=time<arrival;

event=[event(b4arrival) 1 event(˜b4arrival)];

time=[time(b4arrival) arrival time(˜b4arrival)];

if n<c %call admitted

admits=admits+1;

n=n+1;

depart=timenow+exponentialrv(mu,1);

b4depart=time<depart;

event=[event(b4depart) -1 event(˜b4depart)];

time=[time(b4depart) depart time(˜b4depart)];

else

blocks=blocks+1; %one more block, immed departure

disp(sprintf(’Time %10.3d Admits %10d Blocks %10d’,...

timenow,admits,blocks));

end

elseif (eventnow==-1) %departure

n=n-1;

end

end

Figure 5: Discrete event simulation of the blocking switch of Quiz 10.13.

66

Quiz Solutions – Chapter 11

Quiz 11.1

By Theorem 11.2,

µ

Y

= µ

X

∞

−∞

h(t )dt = 2

∞

0

e

−t

dt = 2 (1)

Since R

X

(τ) = δ(τ), the autocorrelation function of the output is

R

Y

(τ) =

∞

−∞

h(u)

∞

−∞

h(v)δ(τ +u −v) dv du =

∞

−∞

h(u)h(τ +u) du (2)

For τ > 0, we have

R

Y

(τ) =

∞

0

e

−u

e

−τ−u

du = e

−τ

∞

0

e

−2u

du =

1

2

e

−τ

(3)

For τ < 0, we can deduce that R

Y

(τ) =

1

2

e

−|τ|

by symmetry. Just to be safe though, we

can double check. For τ < 0,

R

Y

(τ) =

∞

−τ

h(u)h(τ +u) du =

∞

−τ

e

−u

e

−τ−u

du =

1

2

e

τ

(4)

Hence,

R

Y

(τ) =

1

2

e

−|τ|

(5)

Quiz 11.2

The expected value of the output is

µ

Y

= µ

X

∞

¸

n=−∞

h

n

= 0.5(1 +−1) = 0 (1)

The autocorrelation of the output is

R

Y

[n] =

1

¸

i =0

1

¸

j =0

h

i

h

j

R

X

[n +i − j ] (2)

= 2R

X

[n] − R

X

[n −1] − R

X

[n +1] =

¸

1 n = 0

0 otherwise

(3)

Since µ

Y

= 0, The variance of Y

n

is Var[Y

n

] = E[Y

2

n

] = R

Y

[0] = 1.

67

−15 −10 −5 0 5 10 15

0

0.2

0.4

0.6

f

S

X

(

f

)

−1500−1000 −500 0 500 1000 1500

0

2

4

6

8

x 10

f

S

X

(

f

)

−0.2 −0.1 0 0.1 0.2

−5

0

5

10

τ

R

X

(

τ

)

−2 −1 0 1 2

x 10

−3

−5

0

5

10

τ

R

X

(

τ

)

(a) W = 10 (b) W = 1000

Figure 6: The autocorrelation R

X

(τ) and power spectral density S

X

( f ) for process X(t ) in

Quiz 11.5.

Quiz 11.3

By Theorem 11.8, Y =

¸

Y

33

Y

34

Y

35

¸

**is a Gaussian random vector since X
**

n

is

a Gaussian random process. Moreover, by Theorem 11.5, each Y

n

has expected value

E[Y

n

] = µ

X

¸

∞

n=−∞

h

n

= 0. Thus E[Y] = 0. Fo ﬁnd the PDF of the Gaussian vector

Y, we need to ﬁnd the covariance matrix C

Y

, which equals the correlation matrix R

Y

since

Y has zero expected value. One way to ﬁnd the R

Y

is to observe that R

Y

has the Toeplitz

structure of Theorem 11.6 and to use Theorem 11.5 to ﬁnd the autocorrelation function

R

Y

[n] =

∞

¸

i =−∞

∞

¸

j =−∞

h

i

h

j

R

X

[n +i − j ]. (1)

Despite the fact that R

X

[k] is an impulse, using Equation (1) is surprisingly tedious because

we still need to sum over all i and j such that n +i − j = 0.

In this problem, it is simpler to observe that Y = HX where

X =

¸

X

30

X

31

X

32

X

33

X

34

X

35

¸

(2)

and

H =

1

4

⎡

⎣

1 1 1 1 0 0

0 1 1 1 1 0

0 0 1 1 1 1

⎤

⎦

. (3)

In this case, following Theorem 11.7, or by directly applying Theorem 5.13 with µ

X

= 0

and A = H, we obtain R

Y

= HR

X

H

. Since R

X

[n] = δ

n

, R

X

= I, the identity matrix.

68

Thus

C

Y

= R

Y

= HH

=

1

16

⎡

⎣

4 3 2

3 4 3

2 3 4

⎤

⎦

. (4)

It follows (very quickly if you use MATLAB for 3 ×3 matrix inversion) that

C

−1

Y

= 16

⎡

⎣

7/12 −1/2 1/12

−1/2 1 −1/2

1/12 −1/2 7/12

⎤

⎦

. (5)

Thus, the PDF of Y is

f

Y

(y) =

1

(2π)

3/2

[det (C

Y

)]

1/2

exp

−

1

2

y

C

−1

Y

y

. (6)

A disagreeable amount of algebra will show det(C

Y

) = 3/1024 and that the PDF can be

“simpliﬁed” to

f

Y

(y) =

16

√

6π

3

exp

¸

−8

7

12

y

2

33

+ y

2

34

+

7

12

y

2

35

− y

33

y

34

+

1

6

y

33

y

35

− y

34

y

35

¸

. (7)

Equation (7) shows that one of the nicest features of the multivariate Gaussian distribution

is that y

C

−1

Y

y is a very concise representation of the cross-terms in the exponent of f

Y

(y).

Quiz 11.4

This quiz is solved using Theorem 11.9 for the case of k = 1 and M = 2. In this case,

X

n

=

¸

X

n−1

X

n

¸

and

R

X

n

=

¸

R

X

[0] R

X

[1]

R

X

[1] R

X

[0]

¸

=

¸

1.1 0.9

0.9 1.1

¸

(1)

and

R

X

n

X

n+1

= E

¸¸

X

n−1

X

n

¸

X

n+1

¸

=

¸

R

X

[2]

R

X

[1]

¸

=

¸

0.81

0.9

¸

. (2)

The MMSE linear ﬁrst order ﬁlter for predicting X

n+1

at time n is the ﬁlter h such that

←−

h = R

−1

X

n

R

X

n

X

n+1

=

¸

1.1 0.9

0.9 1.1

¸

−1

¸

0.81

0.9

¸

=

1

400

¸

81

261

¸

. (3)

It follows that the ﬁlter is h =

¸

261/400 81/400

¸

**and the MMSE linear predictor is
**

ˆ

X

n+1

=

81

400

X

n−1

+

261

400

X

n

. (4)

to ﬁnd the mean square error, one approach is to follow the method of Example 11.13 and

to directly calculate

e

∗

L

= E

¸

(X

n+1

−

ˆ

X

n+1

)

2

¸

. (5)

69

This method is workable for this simple problem but becomes increasingly tedious for

higher order ﬁlters. Instead, we can derive the mean square error for an arbitary prediction

ﬁlter h. Since

ˆ

X

n+1

=

←−

h

X

n

,

e

∗

L

= E

¸

X

n+1

−

←−

h

X

n

2

¸

(6)

= E

¸

(X

n+1

−

←−

h

X

n

)(X

n+1

−

←−

h

X

n

)

¸

(7)

= E

¸

(X

n+1

−

←−

h

X

n

)(X

n+1

−X

n

←−

h )

¸

(8)

After a bit of algebra, we obtain

e

∗

L

= R

X

[0] −2

←−

h

R

X

n

X

n+1

+

←−

h

R

X

n

←−

h (9)

(10)

with the substitution

←−

h = R

−1

X

n

R

X

n

X

n+1

, we obtain

e

∗

L

= R

X

[0] −R

X

n

X

n+1

R

−1

X

n

R

X

n

X

n+1

(11)

= R

X

[0] −

←−

h

R

X

n

X

n+1

(12)

Note that this is essentially the same result as Theorem 9.7 with Y = X

n

, X = X

n+1

and

ˆ a

=

←−

h

. It is noteworthy that the result is derived in a much simpler way in the proof of

Theorem 9.7 by using the orthoginality property of the LMSE estimator.

In any case, the mean square error is

e

∗

L

= R

X

[0] −

←−

h

R

X

n

X

n+1

= 1.1 −

1

400

¸

81 261

¸

¸

0.81

0.9

¸

=

506

1451

= 0.3487. (13)

recalling that the blind estimate would yield a mean square error of Var[X] = 1.1, we see

that observing X

n−1

and X

n

improves the accuracy of our prediction of X

n+1

.

Quiz 11.5

(1) By Theorem 11.13(b), the average power of X(t ) is

E

¸

X

2

(t )

¸

=

∞

−∞

S

X

( f ) d f =

W

−W

5

W

d f = 10 Watts (1)

(2) The autocorrelation function is the inverse Fourier transform of S

X

( f ). Consulting

Table 11.1, we note that

S

X

( f ) = 10

1

2W

rect

f

2W

(2)

It follows that the inverse transform of S

X

( f ) is

R

X

(τ) = 10 sinc(2Wτ) = 10

sin(2πWτ)

2πWτ

(3)

(3) For W = 10 Hz and W = 1 kHZ, graphs of S

X

( f ) and R

X

(τ) appear in Figure 6.

70

Quiz 11.6

In a sampled system, the discrete time impulse δ[n] has a ﬂat discrete Fourier transform.

That is, if R

X

[n] = 10δ[n], then

S

X

(φ) =

∞

¸

n=−∞

10δ[n]e

−j 2πφn

= 10 (1)

Thus, R

X

[n] = 10δ[n]. (This quiz is really lame!)

Quiz 11.7

Since Y(t ) = X(t −t

0

),

R

XY

(t, τ) = E [X(t )Y(t +τ)] = E [X(t )X(t +τ −t

0

)] = R

X

(τ −t

0

) (1)

We see that R

XY

(t, τ) = R

XY

(τ) = R

X

(τ − t

0

). From Table 11.1, we recall the prop-

erty that g(τ − τ

0

) has Fourier transform G( f )e

−j 2π f τ

0

. Thus the Fourier transform of

R

XY

(τ) = R

X

(τ −t

0

) = g(τ −t

0

) is

S

XY

( f ) = S

X

( f )e

−j 2π f t

0

. (2)

Quiz 11.8

We solve this quiz using Theorem 11.17. First we need some preliminary facts. Let

a

0

= 5,000 so that

R

X

(τ) =

1

a

0

a

0

e

−a

0

|τ|

. (1)

Consulting with the Fourier transforms in Table 11.1, we see that

S

X

( f ) =

1

a

0

2a

2

0

a

2

0

+(2π f )

2

=

2a

0

a

2

0

+(2π f )

2

(2)

The RC ﬁlter has impulse response h(t ) = a

1

e

−a

1

t

u(t ), where u(t ) is the unit step function

and a

1

= 1/RC where RC = 10

−4

is the ﬁlter time constant. From Table 11.1,

H( f ) =

a

1

a

1

+ j 2π f

(3)

(1) Theorem 11.17,

S

XY

( f ) = H( f )S

X

( f ) =

2a

0

a

1

[a

1

+ j 2π f ]

¸

a

2

0

+(2π f )

2

¸. (4)

(2) Again by Theorem 11.17,

S

Y

( f ) = H

∗

( f )S

XY

( f ) = |H( f )|

2

S

X

( f ). (5)

71

Note that

|H( f )|

2

= H( f )H

∗

( f ) =

a

1

(a

1

+ j 2π f )

a

1

(a

1

− j 2π f )

=

a

2

1

a

2

1

+(2π f )

2

(6)

Thus,

S

Y

( f ) = |H( f )|

2

S

X

( f ) =

2a

0

a

2

1

¸

a

2

1

+(2π f )

2

¸ ¸

a

2

0

+(2π f )

2

¸ (7)

(3) To ﬁnd the average power at the ﬁlter output, we can either use basic calculus and

calculate

∞

−∞

S

Y

( f ) d f directly or we can ﬁnd R

Y

(τ) as an inverse transform of

S

Y

( f ). Using partial fractions and the Fourier transform table, the latter method is

actually less algebra. In particular, some algebra will show that

S

Y

( f ) =

K

0

a

2

0

+(2π f )

2

+

K

1

a

1

+(2π f )

2

(8)

where

K

0

=

2a

0

a

2

1

a

2

1

−a

2

0

, K

1

=

−2a

0

a

2

1

a

2

1

−a

2

0

. (9)

Thus,

S

Y

( f ) =

K

0

2a

2

0

2a

2

0

a

2

0

+(2π f )

2

+

K

1

2a

2

1

2a

2

1

a

1

+(2π f )

2

. (10)

Consulting with Table 11.1, we see that

R

Y

(τ) =

K

0

2a

2

0

a

0

e

−a

0

|τ|

+

K

1

2a

2

1

a

1

e

−a

1

|τ|

(11)

Substituting the values of K

0

and K

1

, we obtain

R

Y

(τ) =

a

2

1

e

−a

0

|τ|

−a

0

a

1

e

−a

1

|τ|

a

2

1

−a

2

0

. (12)

The average power of the Y(t ) process is

R

Y

(0) =

a

1

a

1

+a

0

=

2

3

. (13)

Note that the input signal has average power R

X

(0) = 1. Since the RC ﬁlter has a 3dB

bandwidth of 10,000 rad/sec and the signal X(t ) has most of its its signal energy below

5,000 rad/sec, the output signal has almost as much power as the input.

72

Quiz 11.9

This quiz implements an example of Equations (11.146) and (11.147) for a system in

which we ﬁlter Y(t ) = X(t ) + N(t ) to produce an optimal linear estimate of X(t ). The

solution to this quiz is just to ﬁnd the ﬁlter

ˆ

H( f ) using Equation (11.146) and to calculate

the mean square error e

L

∗ using Equation (11.147).

Comment: Since the text omitted the derivations of Equations (11.146) and (11.147), we

note that Example 10.24 showed that

R

Y

(τ) = R

X

(τ) + R

N

(τ), R

Y X

(τ) = R

X

(τ). (1)

Taking Fourier transforms, it follows that

S

Y

( f ) = S

X

( f ) + S

N

( f ), S

Y X

( f ) = S

X

( f ). (2)

Now we can go on to the quiz, at peace with the derivations.

(1) Since µ

N

= 0, R

N

(0) = Var[N] = 1. This implies

R

N

(0) =

∞

−∞

S

N

( f ) d f =

B

−B

N

0

d f = 2N

0

B (3)

Thus N

0

= 1/(2B). Because the noise process N(t ) has constant power R

N

(0) = 1,

decreasing the single-sided bandwidth B increases the power spectral density of the

noise over frequencies | f | < B.

(2) Since R

X

(τ) = sinc(2Wτ), where W = 5,000 Hz, we see from Table 11.1 that

S

X

( f ) =

1

10

4

rect

f

10

4

. (4)

The noise power spectral density can be written as

S

N

( f ) = N

0

rect

f

2B

=

1

2B

rect

f

2B

, (5)

From Equation (11.146), the optimal ﬁlter is

ˆ

H( f ) =

S

X

( f )

S

X

( f ) + S

N

( f )

=

1

10

4

rect

f

10

4

1

10

4

rect

f

10

4

+

1

2B

rect

f

2B

. (6)

73

(3) We produce the output

ˆ

X(t ) by passing the noisy signal Y(t ) through the ﬁlter

ˆ

H( f ).

From Equation (11.147), the mean square error of the estimate is

e

∗

L

=

∞

−∞

S

X

( f )S

N

( f )

S

X

( f ) + S

N

( f )

d f (7)

=

∞

−∞

1

10

4

rect

f

10

4

1

2B

rect

f

2B

1

10

4

rect

f

10

4

+

1

2B

rect

f

2B

d f. (8)

To evaluate the MSE e

∗

L

, we need to whether B ≤ W. Since the problem asks us to

ﬁnd the largest possible B, let’s suppose B ≤ W. We can go back and consider the

case B > W later. When B ≤ W, the MSE is

e

∗

L

=

B

−B

1

10

4

1

2B

1

10

4

+

1

2B

d f =

1

10

4

1

10

4

+

1

2B

=

1

1 +

5,000

B

(9)

To obtain MSE e

∗

L

≤ 0.05 requires B ≤ 5,000/19 = 263.16 Hz.

Although this completes the solution to the quiz, what is happening may not be obvious.

The noise power is always Var[N] = 1 Watt, for all values of B. As B is decreased, the PSD

S

N

( f ) becomes increasingly tall, but only over a bandwidth B that is decreasing. Thus as

B descreases, the ﬁlter

ˆ

H( f ) makes an increasingly deep and narrow notch at frequencies

| f | ≤ B. Two examples of the ﬁlter

ˆ

H( f ) are shown in Figure 7. As B shrinks, the ﬁlter

suppresses less of the signal of X(t ). The result is that the MSE goes down.

Finally, we note that we can choose B very large and also achieve MSE e

∗

L

= 0.05. In

particular, when B > W = 5000, S

N

( f ) = 1/2B over frequencies | f | < W. In this case,

the Wiener ﬁlter

ˆ

H( f ) is an ideal (ﬂat) lowpass ﬁlter

ˆ

H( f ) =

⎧

⎨

⎩

1

10

4

1

10

4

+

1

2B

| f | < 5,000,

0 otherwise.

(10)

Thus increasing B spreads the constant 1 watt of power of N(t ) over more bandwidth. The

Wiener ﬁlter removes the noise that is outside the band of the desired signal. The mean

square error is

e

∗

L

=

5000

−5000

1

10

4

1

2B

1

10

4

+

1

2B

d f =

1

2B

1

10

4

+

1

2B

=

1

B

5000

+1

(11)

In this case, B ≥ 9.5 ×10

4

guarantees e

∗

L

≤ 0.05.

Quiz 11.10

It is fairly straightforward to ﬁnd S

X

(φ) and S

Y

(φ). The only thing to keep in mind is

to use fftc to transform the autocorrelation R

X

[ f ] into the power spectral density S

X

(φ).

The following MATLAB program generates and plots the functions shown in Figure 8

74

−5000 −2000 0 2000 5000

0

0.5

1

f

H

(

f

)

−5000 −2000 0 2000 5000

0

0.5

1

f

H

(

f

)

B = 500 B = 2500

Figure 7: Wiener ﬁlter for Quiz 11.9.

%mquiz11.m

N=32;

rx=[2 4 2]; SX=fftc(rx,N); %autocorrelation and PSD

stem(0:N-1,abs(sx));

xlabel(’n’);ylabel(’S_X(n/N)’);

h2=0.5*[1 1]; H2=fft(h2,N); %impulse/filter response: M=2

SY2=SX.* ((abs(H2)).ˆ2);

figure; stem(0:N-1,abs(SY2)); %PSD of Y for M=2

xlabel(’n’);ylabel(’S_{Y_2}(n/N)’);

h10=0.1*ones(1,10); H10=fft(h10,N); %impulse/filter response: M=10

SY10=sx.*((abs(H10)).ˆ2);

figure; stem(0:N-1,abs(SY10));

xlabel(’n’);ylabel(’S_{Y_{10}}(n/N)’);

Relative to M = 2, when M = 10, the ﬁlter H(φ) ﬁlters out almost all of the high

frequency components of X(t ). In the context of Example 11.26, the low pass moving

average ﬁlter for M = 10 removes the high frquency components and results in a ﬁlter

output that varies very slowly.

As an aside, note that the vectors SX, SY2 and SY10 in mquiz11 should all be real-

valued vectors. However, the ﬁnite numerical precision of MATLAB results in tiny imagi-

nary parts. Although these imaginary parts have no computational signiﬁcance, they tend

to confuse the stem function. Hence, we generate stem plots of the magnitude of each

power spectral density.

75

0 5 10 15 20 25 30 35

0

5

10

n

S

X

(

n

/

N

)

0 5 10 15 20 25 30 35

0

5

10

n

S

Y

2

(

n

/

N

)

0 5 10 15 20 25 30 35

0

5

10

n

S

Y

1

0

(

n

/

N

)

Figure 8: For Quiz 11.10, graphs of S

X

(φ), S

Y

(n/N) for M = 2, and S

φ

(n/N) for M = 10

using an N = 32 point DFT.

76

Quiz Solutions – Chapter 12

Quiz 12.1

The system has two states depending on whether the previous packet was received in

error. From the problem statement, we are given the conditional probabilities

P

¸

X

n+1

= 0|X

n

= 0

¸

= 0.99 P

¸

X

n+1

= 1|X

n

= 1

¸

= 0.9 (1)

Since each X

n

must be either 0 or 1, we can conclude that

P

¸

X

n+1

= 1|X

n

= 0

¸

= 0.01 P

¸

X

n+1

= 0|X

n

= 1

¸

= 0.1 (2)

These conditional probabilities correspond to the transition matrix and Markov chain:

0 1

0.01

0.1

0.99 0.9

P =

¸

0.99 0.01

0.10 0.90

¸

(3)

Quiz 12.2

From the problem statement, the Markov chain and the transition matrix are

0 1 1

0.6 0.2

0.2 0.6

0.4 0.6 0.4

P =

⎡

⎣

0.4 0.6 0

0.2 0.6 0.2

0 0.6 0.4

⎤

⎦

(1)

The eigenvalues of P are

λ

1

= 0 λ

2

= 0.4 λ

3

= 1 (2)

We can diagonalize P into

P = S

−1

DS =

⎡

⎣

−0.6 0.5 1

0.4 0 1

−0.6 −0.5 1

⎤

⎦

⎡

⎣

λ

1

0 0

0 λ

2

0

0 0 λ

3

⎤

⎦

⎡

⎣

−0.5 1 −0.5

1 0 −1

0.2 0.6 0.2

⎤

⎦

(3)

where s

i

, the i th row of S, is the left eigenvector of P satisfying s

i

P = λ

i

s

i

. Algebra will

verify that the n-step transition matrix is

P

n

= S

−1

D

n

S =

⎡

⎣

0.2 0.6 0.2

0.2 0.6 0.2

0.2 0.6 0.2

⎤

⎦

+(0.4)

n

⎡

⎣

0.5 0 −0.5

0 0 0

−0.5 0 0.5

⎤

⎦

(4)

Quiz 12.3

The Markov chain describing the factory status and the corresponding state transition

matrix are

77

2

0 1

0.9

0.1

1

1

P =

⎡

⎣

0.9 0.1 0

0 0 1

1 0 0

⎤

⎦

(1)

With π =

¸

π

0

π

1

π

2

¸

, the system of equations π

= π

P yields π

1

= 0.1π

0

and

π

2

= π

1

. This implies

π

0

+π

1

+π

2

= π

0

(1 +0.1 +0.1) = 1 (2)

It follows that the limiting state probabilities are

π

0

= 5/6, π

1

= 1/12, π

2

= 1/12. (3)

Quiz 12.4

The communicating classes are

C

1

= {0, 1} C

2

= {2, 3} C

3

= {4, 5, 6} (1)

The states in C

1

and C

3

are aperiodic. The states in C

2

have period 2. Once the system

enters a state in C

1

, the class C

1

is never left. Thus the states in C

1

are recurrent. That

is, C

1

is a recurrent class. Similarly, the states in C

3

are recurrent. On the other hand, the

states in C

2

are transient. Once the system exits C

2

, the states in C

2

are never reentered.

Quiz 12.5

At any time t , the state n can take on the values 0, 1, 2, . . .. The state transition proba-

bilities are

P

n−1,n

= P [K > n|K > n −1] =

P [K > n]

P [K > n −1]

(1)

P

n−1,0

= P [K = n|K > n −1] =

P [K = n]

P [K > n −1]

(2)

(3)

The Markov chain resembles

0 1

P K=2 [ ]

P K= [ 1]

3 4

P K=4 [ ]

2

P K=3 [ ]

P K=5 [ ]

1 1 1 1 1

… ...

78

The stationary probabilities satisfy

π

0

= π

0

P [K = 1] +π

1

, (4)

π

1

= π

0

P [K = 2] +π

2

, (5)

.

.

.

π

k−1

= π

0

P [K = k] +π

k

, k = 1, 2, . . . (6)

From Equation (4), we obtain

π

1

= π

0

(1 − P [K = 1]) = π

0

P [K > 1] (7)

Similarly, Equation (5) implies

π

2

= π

1

−π

0

P [K = 2] = π

0

(P [K > 1] − P [K = 2]) = π

0

P [K > 2] (8)

This suggests that π

k

= π

0

P[K > k]. We verify this pattern by showing that π

k

=

π

0

P[K > k] satisﬁes Equation (6):

π

0

P [K > k −1] = π

0

P [K = k] +π

0

P [K > k] . (9)

When we apply

¸

∞

k=0

π

k

= 1, we obtain π

0

¸

∞

n=0

P[K > k] = 1. From Problem 2.5.11,

we recall that

¸

∞

k=0

P[K > k] = E[K]. This implies

π

n

=

P [K > n]

E [K]

(10)

This Markov chain models repeated random countdowns. The system state is the time until

the counter expires. When the counter expires, the system is in state 0, and we randomly

reset the counter to a new value K = k and then we count down k units of time. Since we

spend one unit of time in each state, including state 0, we have k −1 units of time left after

the state 0 counter reset. If we have a random variable W such that the PMF of W satisﬁes

P

W

(n) = π

n

, then W has a discrete PMF representing the remaining time of the counter at

a time in the distant future.

Quiz 12.6

(1) By inspection, the number of transitions need to return to state 0 is always a multiple

of 2. Thus the period of state 0 is d = 2.

(2) To ﬁnd the stationary probabilities, we solve the system of equations π = πP and

¸

3

i =0

π

i

= 1:

π

0

= (3/4)π

1

+(1/4)π

3

(1)

π

1

= (1/4)π

0

+(1/4)π

2

(2)

π

2

= (1/4)π

1

+(3/4)π

3

(3)

1 = π

0

+π

1

+π

2

+π

3

(4)

79

Solving the second and third equations for π

2

and π

3

yields

π

2

= 4π

1

−π

0

π

3

= (4/3)π

2

−(1/3)π

1

= 5π

1

−(4/3)π

0

(5)

Substituting π

3

back into the ﬁrst equation yields

π

0

= (3/4)π

1

+(1/4)π

3

= (3/4)π

1

+(5/4)π

1

−(1/3)π

0

(6)

This implies π

1

= (2/3)π

0

. It follows from the ﬁrst and second equations that

π

2

= (5/3)π

0

and π

3

= 2π

0

. Lastly, we choose π

0

so the state probabilities sum to

1:

1 = π

0

+π

1

+π

2

+π

3

= π

0

1 +

2

3

+

5

3

+2

=

16

3

π

0

(7)

It follows that the state probabilities are

π

0

=

3

16

π

1

=

2

16

π

2

=

5

16

π

3

=

6

16

(8)

(3) Since the system starts in state 0 at time 0, we can use Theorem 12.14 to ﬁnd the

limiting probability that the system is in state 0 at time nd:

lim

n→∞

P

00

(nd) = dπ

0

=

3

8

(9)

Quiz 12.7

The Markov chain has the same structure as that in Example 12.22. The only difference

is the modiﬁed transition rates:

0 1

1

3 4

( ) 2/3

a

1 - ( ) 2/3

a

( ) 3/4

a

1 - 3/4 ( )

a

( ) 4/5

a

1 - 4/5 ( )

a

2

( ) 1/2

a

1- 1/2 ( )

a

…

The event T

00

> n occurs if the system reaches state n before returning to state 0, which

occurs with probability

P [T

00

> n] = 1 ×

1

2

α

×

2

3

α

×· · · ×

n −1

n

α

=

1

n

α

. (1)

Thus the CDF of T

00

satisﬁes F

T

00

(n) = 1−P[T

00

> n] = 1−1/n

α

. To determine whether

state 0 is recurrent, we observe that for all α > 0

P [V

00

] = lim

n→∞

F

T

00

(n) = lim

n→∞

1 −

1

n

α

= 1. (2)

80

Thus state 0 is recurrent for all α > 0. Since the chain has only one communicating class,

all states are recurrent. ( We also note that if α = 0, then all states are transient.)

To determine whether the chain is null recurrent or positive recurrent, we need to calcu-

late E[T

00

]. In Example 12.24, we did this by deriving the PMF P

T

00

(n). In this problem,

it will be simpler to use the result of Problem 2.5.11 which says that

¸

∞

k=0

P[K > k] =

E[K] for any non-negative integer-valued random variable K. Applying this result, the

expected time to return to state 0 is

E [T

00

] =

∞

¸

n=0

P [T

00

> n] = 1 +

∞

¸

n=1

1

n

α

. (3)

For 0 < α ≤ 1, 1/n

α

≥ 1/n and it follows that

E [T

00

] ≥ 1 +

∞

¸

n=1

1

n

= ∞. (4)

We conclude that the Markov chain is null recurrent for 0 < α ≤ 1. On the other hand, for

α > 1,

E [T

00

] = 2 +

∞

¸

n=2

1

n

α

. (5)

Note that for all n ≥ 2

1

n

α

≤

n

n−1

dx

x

α

(6)

This implies

E [T

00

] ≤ 2 +

∞

¸

n=2

n

n−1

dx

x

α

(7)

= 2 +

∞

1

dx

x

α

(8)

= 2 +

x

−α+1

−α +1

∞

1

= 2 +

1

α −1

< ∞ (9)

Thus for all α > 1, the Markov chain is positive recurrent.

Quiz 12.8

The number of customers in the ”friendly” store is given by the Markov chain

1 i i+1

p p p

( )( ) 1-p 1-q ( )( ) 1-p 1-q ( )( ) 1-p 1-q ( )( ) 1-p 1-q

( ) 1-p q ( ) 1-p q ( ) 1-p q ( ) 1-p q

0

××× ×××

81

In the above chain, we note that (1 − p)q is the probability that no new customer arrives,

an existing customer gets one unit of service and then departs the store.

By applying Theorem 12.13 with state space partitioned between S = {0, 1, . . . , i } and

S

**= {i +1, i +2, . . .}, we see that for any state i ≥ 0,
**

π

i

p = π

i +1

(1 − p)q. (1)

This implies

π

i +1

=

p

(1 − p)q

π

i

. (2)

Since Equation (2) holds for i = 0, 1, . . ., we have that π

i

= π

0

α

i

where

α =

p

(1 − p)q

. (3)

Requiring the state probabilities to sum to 1, we have that for α < 1,

∞

¸

i =0

π

i

= π

0

∞

¸

i =0

α

i

=

π

0

1 −α

= 1. (4)

Thus for α < 1, the limiting state probabilities are

π

i

= (1 −α)α

i

, i = 0, 1, 2, . . . (5)

In addition, for α ≥ 1 or, equivalently, p ≥ q/(1 − q), the limiting state probabilities do

not exist.

Quiz 12.9

The continuous time Markov chain describing the processor is

0 1

2

3.01

3 4

2

3

2

3

2

2

3

0.01

0.01

0.01

Note that q

10

= 3.1 since the task completes at rate 3 per msec and the processor reboots

at rate 0.1 per msec and the rate to state 0 is the sum of those two rates. From the Markov

chain, we obtain the following useful equations for the stationary distribution.

5.01p

1

= 2p

0

+3p

2

5.01p

2

= 2p

1

+3p

3

5.01p

3

= 2p

2

+3p

4

3.01p

4

= 2p

3

We can solve these equations by working backward and solving for p

4

in terms of p

3

, p

3

in terms of p

2

and so on, yielding

p

4

=

20

31

p

3

p

3

=

620

981

p

2

p

2

=

19620

31431

p

1

p

1

=

628, 620

1, 014, 381

p

0

(1)

82

Applying p

0

+ p

1

+ p

2

+ p

3

+ p

4

= 1 yields p

0

= 1, 014, 381/2, 443, 401 and the

stationary probabilities are

p

0

= 0.4151 p

1

= 0.2573 p

2

= 0.1606 p

3

= 0.1015 p

4

= 0.0655 (2)

Quiz 12.10

The M/M/c/∞queue has Markov chain

c c+1 1 0

λ λ λ λ λ

µ 2µ

cµ cµ cµ

From the Markov chain, the stationary probabilities must satisfy

p

n

=

¸

(ρ/n) p

n−1

n = 1, 2, . . . , c

(ρ/c) p

n−1

n = c +1, c +2, . . .

(1)

It is straightforward to show that this implies

p

n

=

¸

p

0

ρ

n

/n! n = 1, 2, . . . , c

p

0

(ρ/c)

n−c

ρ

c

/c! n = c +1, c +2, . . .

(2)

The requirement that

¸

∞

n=0

p

n

= 1 yields

p

0

=

c

¸

n=0

ρ

n

/n! +

ρ

c

c!

ρ/c

1 −ρ/c

−1

(3)

83

**Quiz Solutions – Chapter 1
**

Quiz 1.1 In the Venn diagrams for parts (a)-(g) below, the shaded area represents the indicated set.

M T O M T O M T O

(1) R = T c

(2) M ∪ O

(3) M ∩ O

M T

O

M T

O

M T

O

(4) R ∪ M Quiz 1.2 (1) A1 = {vvv, vvd, vdv, vdd} (2) B1 = {dvv, dvd, ddv, ddd} (3) A2 = {vvv, vvd, dvv, dvd} (4) B2 = {vdv, vdd, ddv, ddd} (5) A3 = {vvv, ddd} (6) B3 = {vdv, dvd}

(4) R ∩ M

(6) T c − M

(7) A4 = {vvv, vvd, vdv, dvv, vdd, dvd, ddv} (8) B4 = {ddd, ddv, dvd, vdd} Recall that Ai and Bi are collectively exhaustive if Ai ∪ Bi = S. Also, Ai and Bi are mutually exclusive if Ai ∩ Bi = φ. Since we have written down each pair Ai and Bi above, we can simply check for these properties. The pair A1 and B1 are mutually exclusive and collectively exhaustive. The pair A2 and B2 are mutually exclusive and collectively exhaustive. The pair A3 and B3 are mutually exclusive but not collectively exhaustive. The pair A4 and B4 are not mutually exclusive since dvd belongs to A4 and B4 . However, A4 and B4 are collectively exhaustive. 2

Quiz 1.3 There are exactly 50 equally likely outcomes: s51 through s100 . Each of these outcomes has probability 0.02. (1) P[{s79 }] = 0.02 (2) P[{s100 }] = 0.02 (3) P[A] = P[{s90 , . . . , s100 }] = 11 × 0.02 = 0.22 (4) P[F] = P[{s51 , . . . , s59 }] = 9 × 0.02 = 0.18 (5) P[T ≥ 80] = P[{s80 , . . . , s100 }] = 21 × 0.02 = 0.42 (6) P[T < 90] = P[{s51 , s52 , . . . , s89 }] = 39 × 0.02 = 0.78 (7) P[a C grade or better] = P[{s70 , . . . , s100 }] = 31 × 0.02 = 0.62 (8) P[student passes] = P[{s60 , . . . , s100 }] = 41 × 0.02 = 0.82 Quiz 1.4 We can describe this experiment by the event space consisting of the four possible events V B, V L, D B, and DL. We represent these events in the table: V D L 0.35 ? B ? ? In a roundabout way, the problem statement tells us how to ﬁll in the table. In particular, P [V ] = 0.7 = P [V L] + P [V B] P [L] = 0.6 = P [V L] + P [DL] (1) (2)

Since P[V L] = 0.35, we can conclude that P[V B] = 0.35 and that P[DL] = 0.6 − 0.35 = 0.25. This allows us to ﬁll in two more table entries: V D L 0.35 0.25 B 0.35 ? The remaining table entry is ﬁlled in by observing that the probabilities must sum to 1. This implies P[D B] = 0.05 and the complete table is V D L 0.35 0.25 B 0.35 0.05 Finding the various probabilities is now straightforward: 3

(1) P[DL] = 0.25 (2) P[D ∪ L] = P[V L] + P[DL] + P[D B] = 0.35 + 0.25 + 0.05 = 0.65. (3) P[V B] = 0.35 (4) P[V ∪ L] = P[V ] + P[L] − P[V L] = 0.7 + 0.6 − 0.35 = 0.95 (5) P[V ∪ D] = P[S] = 1 (6) P[L B] = P[L L c ] = 0 Quiz 1.5 (1) The probability of exactly two voice calls is P [N V = 2] = P [{vvd, vdv, dvv}] = 0.3 (2) The probability of at least one voice call is P [N V ≥ 1] = P [{vdd, dvd, ddv, vvd, vdv, dvv, vvv}] = 6(0.1) + 0.2 = 0.8 An easier way to get the same answer is to observe that P [N V ≥ 1] = 1 − P [N V < 1] = 1 − P [N V = 0] = 1 − P [{ddd}] = 0.8 (4) (2) (3) (1)

(3) The conditional probability of two voice calls followed by a data call given that there were two voice calls is 1 P [{vvd} , N V = 2] P [{vvd}] 0.1 = (5) = = P [{vvd} |N V = 2] = P [N V = 2] P [N V = 2] 0.3 3 (4) The conditional probability of two data calls followed by a voice call given there were two voice calls is P [{ddv} , N V = 2] P [{ddv} |N V = 2] = =0 (6) P [N V = 2] The joint event of the outcome ddv and exactly two voice calls has probability zero since there is only one voice call in the outcome ddv. (5) The conditional probability of exactly two voice calls given at least one voice call is P [N V = 2, N V ≥ 1] P [N V = 2] 0.3 3 = = = (7) P [N V = 2|Nv ≥ 1] = P [N V ≥ 1] P [N V ≥ 1] 0.8 8 (6) The conditional probability of at least one voice call given there were exactly two voice calls is P [N V ≥ 1, N V = 2] P [N V = 2] P [N V ≥ 1|N V = 2] = = =1 (8) P [N V = 2] P [N V = 2] Given that there were two voice calls, there must have been at least one voice call. 4

(2) The probability of the joint event is P [N V ≥ 1. (1) First.68) = 0.96) = P [N V = 2.8) = 0.544.8) = 0.16 P[{vd}] = (0. the events are dependent. we observe that P [N V ≥ 1] = P [{vd. Using the probabilities of the outcomes. each event has probability P [C2 = v] = P [{dv. vv}] = 0.64 Also. P [N V is even] = P [{dd.8)2 = 0. we conﬁrm that the events are independent.2) = 0.16.16 P[{dd}] = (0. P[C1 = v] = 0. vv}] = 0.96 Finally.2)(0.64 P[{dv}] = (0. C2 = v] = P [{dv}] = 0. Just to be sure. we make the comparison P [N V = 2] P [N V ≥ 1] = (0.8 so that P [N V ≥ 1] P [C1 = v] = (0.96)(0. the events are dependent. {C1 = d} are independent events.6 In this experiment.80 From part (a). {C2 = v}.2)(0.8)(0. N V is even] = 0. we now can test for the independence of events. N V ≥ 1] = P [N V = 2] = P [{vv}] = 0. 5 (7) (5) (4) (3) (2) (1)
.64 Next.16 (6) Since P[C1 = d]P[C2 = v] = (0. P[N V ≥ 1] = 0.Quiz 1. we can do the calculations to check: P [C1 = d.64)(0. N V is even] = P [{vv}] = 0.8)(0. Further.04
When checking the independence of any two events A and B. it’s wise to avoid intuition and simply check whether P[AB] = P[A]P[B].68 (8) Thus.8.8) = 0. Since P[C2 = v. C1 = v] Hence. there are four outcomes with probabilities P[{vv}] = (0. and the ﬁrst call is a data call. (3) The problem statement that the calls were independent implies that the events the second call is a voice call. (4) The probability of the joint event is P [C2 = v.96.544. vv}] = 0. N V ≥ 1] which shows the two events are dependent. P[C2 = v]P[N V is even] = (0. Note that this shouldn’t be surprising since we used the information that the calls were independent in the problem statement to determine the probabilities of the outcomes. dv. we calculate the probability of the joint event: P [N V = 2.768 = P [N V ≥ 1. vv}] = 0. C1 = v] = P [{vd.2)2 = 0.

In this case. The tree for the experiment is
0.2
The user is found unless all three paging attempts fail.9 (1) In this problem. There are 4 = 6 ways to do this. (4) For the constant ratio code.8 (1) We can view choosing each bit in the code word as a subexperiment.8 ¨ F1 0.2 0. The failure probability is = 1 − p and the success probability is 1 − = p.8 ¨ F2 0. That is. The number of ways of choosing such N a code word is M . the probability of k bits in error and 100 − k correctly received bits is P Sk. 1001.2 0. Thus by the fundamental principle of counting. k bits received in error is the same as k failures in 100 trials. it is also possible to simply enumerate the six code words: 1100. there are six code words with exactly two zeroes. (3) When the ﬁrst bit must be a zero.100−k = 100 k 6
k
(1 − )100−k
(1)
. 0110. 2 For this problem. we have two choices. For each of the next three bits. Thus the probability the user is found is c c c P [F] = 1 − P F1 F2 F3 = 1 − (0. we can specify a code word by choosing M of the bits to be ones. there are 1 × 2 × 2 × 2 = 8 ways of choosing a code word. The other two bits then must be ones. there are 2 × 2 × 2 × 2 = 24 = 16 possible code words. 0011.992 (1) Quiz 1. 3 Quiz 1. (2) An experiment that can yield all possible code words with two zeroes is to choose which 2 bits (out of 4 bits) will be zero. Hence. The other N − M bits will be zeroes. Each subexperiment has two possible outcomes: 0 and 1.Quiz 1.7 Let Fi denote the event that that the user is found on page i. then the ﬁrst subexperiment of choosing the ﬁrst bit has only one outcome. 1010.2)3 = 0. there are 8 = 56 code words. 0101. For N = 8 and M = 3.8 ¨ F3 ¨¨ ¨¨ ¨¨ ¨ ¨¨ ¨¨ c c c ¨¨ F3 F1 ¨ F2 ¨ 0.

we ﬁrst generate a vector R of 100 random numbers. the transistors in the chip are like devices in series.9 < R(i).1:3)
(1) (2)
(3)
For a M ATLAB simulation. P [C8 ] = The probability a memory module works is P [M] = P [C8 ] + P [C9 ] = p 8n (9 − 8 p n ) Quiz 1.97 = 0. + (2*(R>0. 7
.99 = 100(0. then X(i)=3.. The probability that a chip works is P[C] = pn . we use the hist function to count how many occurences of each possible value of X(i).99) 8 = 0.5 and 0...98 = 4950(0.99) (2) The probability a packet is decoded correctly is just P [C] = P S0.For
= 0.*(R<=0.100 + P S1.11
R=rand(1..01) (0.1849 P S3.4 < R(i) and R(i)<=0.1. we generate vector X as a function of R to represent the 3 possible outcomes of a ﬂip. Lastly. These three cases will have probabilities 0. Y=hist(X. then X(i)=2. X(i)=2 if ﬂip i was tails.01. Let Ck denote the event that exactly k chips work. That is.3700
9 97
P S2. 0.3660 P S1.99)
2 3 99
(2) (3) (4) (5)
= 0. P S0. we note there are three cases:
• If R(i) <= 0. Thus each P[Ck ] has the binomial probability 9 (P [C])8 (1 − P [C])9−8 = 9 p 8n (1 − p n ).97 = 161.9)).100).100 = (1 − )100 = (0. and X(i)=3) is ﬂip i landed on the edge. then X(i)=1.0610
(6)
Quiz 1.4.4). + (3*(R>0. 8 P [C9 ] = (P [C])9 = p 9n . • If 0. To see how this works.4. The module works if either 8 chips work or 9 chips work.99)100 = 0.9)) . Second.4) .98 + P S3.01) (0. X=(R<= 0.9819 = 0. X(i)=1 if ﬂip i was heads.10 Since the chip works only if all n transistors work. • If 0. 700(0.01)(0.9. chip failures are also independent.99 + P S2. Since transistor failures are independent of each other.

1 The sample space. (2) P[N = 1] = PN (1) = c = 6/11 (3) P[N ≥ 2] = PN (2) + PN (3) = c/2 + c/3 = 5/11 (4) P[N > 3] =
∞ n=4
PN (n) = 0
Quiz 2.1. the trial is a success.5 0. probabilities and corresponding grades for the experiment are Outcome P[·] BB BC CB CC Quiz 2.2 (1) To ﬁnd c. .16 2
PN (n) = c 1 +
n=1
1 1 + 2 3
=1
(1)
This implies c = 6/11.3 Decoding each transmitted bit is an independent trial where we call a bit error a “success. we recall that the PMF must sum to 1. then the probability exactly 10 bits are sent is P [X = 10] = PX (10) = (0.24 2. (1) The random variable X is the number of trials up to and including the ﬁrst success. that is.5 0. with probability p.1)(0. 2. the remaining parts are straightforward.24 2. . That is. 0 otherwise (1)
(2) If p = 0.” Each bit is in error. Similar to Example 2.Quiz Solutions – Chapter 2
Quiz 2.11.0 0. .9)9 = 0.
3
G
0.36 3. X has the geometric PMF PX (x) = p(1 − p)x−1 x = 1. Now that we have found c. Now we can interpret each experiment in the generic context of independent trials.0387 8 (2)
.

the probability that the third error occurs on bit 12 is PZ (12) = 11 (0. .4 Each of these probabilities can be read off the CDF FY (y).99)98 = 0. P [X ≥ 10] = P [ﬁrst 10 bits are correct] = (1 − p)10 For p = 0.01)2 (0.1. However.15) PZ (z) = z−1 3 p (1 − p)z−3 2 (9)
Note that PZ (z) > 0 for z = 3. FY (y) takes the upper value FY (y0 ).01. (3) The random variable Y is the number of successes in 100 independent trials.01)2 (0.01)(0.3487.99)99 + = 0. (6) If p = 0. its even easier to observe that X ≥ 10 if the ﬁrst 10 bits are transmitted correctly. This x=10 sum is not too hard to calculate.0645 2 (10)
Quiz 2.The probability that at least 10 bits are sent is P[X ≥ 10] = ∞ PX (x).99)100 + 100(0.99)98 2 (6) (7) (8)
(5) Random variable Z is the number of trials up to and including the third success. 4. 5.25)3 (0.25. However.. Y has the binomial PMF PY (y) = 100 y p (1 − p)100−y y (4) (3)
If p = 0. . we must keep in + mind that when FY (y) has a discontinuity at y0 . Just as in Example 2. Thus Z has the Pascal PMF (see Example 2.75)9 = 0. . (1) P[Y < 1] = FY (1− ) = 0 9
. That is.13.910 = 0. P[X ≥ 10] = 0. the probability of exactly 2 errors is P [Y = 2] = PY (2) = 100 (0.1849 2 (5)
(4) The probability of no more than 2 errors is P [Y ≤ 2] = PY (0) + PY (1) + PY (2) = (0.9207 100 (0.

7.8 = 0.7 c = 25 PC (c) = 0.6 (6) P[Y = 3] = P[Y ≤ 3] − P[Y < 3] = FY (3+ ) − FY (3− ) = 0.8 = 0 Quiz 2.6 (3) P[Y > 2] = 1 − P[Y ≤ 2] = 1 − FY (2) = 1 − 0.8 − 0. Otherwise.1¨¨ ¨ ¨ ¨ 0.1) = 62
(2)
(3) (4)
10
.6 = 0.3 c = 40 (1) ⎩ 0 otherwise (2) The expected value of C is E [C] = 25(0. the cost T is T = 25N + 40(3 − N ) = 120 − 15N (2) To ﬁnd the PMF of T .3) + 120(0.2 (4) P[Y ≥ 2] = 1 − P[Y < 2] = 1 − FY (2− ) = 1 − 0. a call is a voice call and C = 25.5 (1) With probability 0. the expected value of T is E [T ] = 75PT (75) + 90PT (90) + 105PT (105) + 120PT (120) = (75 + 90 + 105)(0. 90.6 (1) As a function of N . we can write down the PMF of T : ⎧ ⎨ 0. with probability 0.3.1 t = 120 ⎩ 0 otherwise From the PMF PT (t).3 t = 75.3 N =2 •T =90 r rr 0.3 N =3 •T =75
From the tree. 105 PT (t) = 0.7) + 40(0. This corresponds to the PMF ⎧ ⎨ 0. we can draw the following tree:
N =0 •T =120 0.4 (5) P[Y = 1] = P[Y ≤ 1] − P[Y < 1] = FY (1+ ) − FY (1− ) = 0. we have a data call and C = 40.3) = 29.3$$N =1 •T =105 $
(2)
(1)
$ $$ ¨¨$ rr rr0.5 cents Quiz 2.(2) P[Y ≤ 1] = FY (1) = 0.

5) = 1.8 The PMF PN (n) allows to calculate each of the desired quantities.2) + 4(0.4
(2)
(3) The variance of N is Var[N ] = E N 2 − (E [N ])2 = 2.4) + 22 (0. the expected number of applications is
4
E [A] =
a=1
a PA (a) = 1(0.3) + 6(0. However.4) + 2(0.4) + 4(0.1) = 4.4 − (1. g(E[A]) = g(2) = 4.4) + 2(0.4)2 = 0.2) + 8(0. Quiz 2. (1) The expected value of N is
2
E [N ] =
n=0
n PN (n) = 0(0. (3)
11
.663.1) + 1(0. The two quantities are different because g(A) is not of the form α A + β. 2 g(A) = 6 A = 3 ⎩ 8 A=4 (3) By Theorem 2.44 (4) The standard deviation is σ N = √ Var[N ] = √ 0.4
(1)
(2) The second moment of N is
2
E N
2
=
n=0
n 2 PN (n) = 02 (0.Quiz 2.7 (1) Using Deﬁnition 2.1) = 2
(1)
(2) The number of memory chips is M = g(A) where ⎧ ⎨ 4 A = 1.8 = g(E[A]).5) = 2. the expected number of memory chips is
4
(2)
E [M] =
a=1
g(A)PA (a) = 4(0.8
(3)
Since E[A] = 2.44 = 0.1) + 12 (0.3) + 3(0.10.14. E[M] = 4.

8. 4.15625 12
.25) n = 1.02 n = 1. From Theorem 1. 3. 3. 2.10 (the law of total probability). 7. we learn that the conditional PMF of N given the event I is 0.Quiz 2. 9.155)(5) + (0. 7. 3.75) + 0. . E [N |N ≤ 10] =
n 5
0
n ≤ 10 otherwise
(7)
(8)
(9)
n PN |N ≤10 (n)
10
(10)
=
n=1
n(0.005/0.005)(5) = 0.02(0. 10 ⎩ 0 otherwise (5) Once we have the conditional PMF.02(0. we ﬁnd the PMF of N is PN (n) = PN |T (n) P [T ] + PN |I (n) P [I ] ⎧ ⎨ 0. . 4. 7. 3.2 n = 1. . 5 0 otherwise (2)
(3) The problem statement tells us that P[T ] = 1 − P[I ] = 3/4. 4.2(0.17. .00625 n = 6. 2. 4. the conditional PMF of N given N ≤ 10 is PN |N ≤10 (n) =
PN (n) P[N ≤10]
⎧ ⎨ 0. . 4. 5 = 0. 3. 2. 50 PN |I (n) = (1) 0 otherwise (2) Also from the problem statement. . . .9 (1) From the problem statement. 9. 10 ⎩ 0 otherwise ⎧ ⎨ 0.19375) +
n=6
n(0.8 n = 1. 50 = 0(0. . 2. .005 n = 6. 7. .155 n = 1.00625)
(11) (12)
= 3. the conditional PMF of N given the event T is PN |T (n) = 0. 5 = 0.25) ⎩ 0 otherwise ⎧ ⎨ 0. 5 = 0.19375 n = 1.155/0.75) + 0.80
(6)
By Theorem 2. 2. 5 n = 6. . calculating conditional expectations is easy. 50 ⎩ 0 otherwise (4) First we ﬁnd
10
(3) (4)
(5)
P [N ≤ 10] =
n=1
PN (n) = (0.8 n = 6. 2. 8.

/K. 2. . m k . K=(1:k)’.15625)2 = 2. What is observed in these ﬁgures is that for small n. X=duniformrv(0.10 The function samplemean(k) generates and plots ﬁve m n sequences for n = 1. for i=1:5.M).71875 The conditional variance is Var[N |N ≤ 10] = E N 2 |N ≤ 10 − (E [N |N ≤ 10])2 = 12. The ith column M(:. . m n is fairly random but as n gets 13
. .10.00625)
n=6
=
n=1
n (0. end.
Examples of the function calls (a) samplemean(100) and (b) samplemean(1000) are shown in Figure 1.71875 − (3. .5). .75684
(16) (17)
Quiz 2. .
function M=samplemean(k). m 2 .i)=cumsum(X). .00625) = 12.10 8 6 4 2 0 0 50 100
10 8 6 4 2 0 0 500 1000
(a) samplemean(100)
(b) samplemean(1000)
Figure 1: Two examples of the output of samplemean(k) (6) To ﬁnd the conditional variance. . Each time samplemean(k) is called produces a random output. we ﬁrst ﬁnd the conditional second moment E N 2 |N ≤ 10 =
n 5
n 2 PN |N ≤10 (n)
10
(13) n 2 (0. k. plot(K.19375) +
2
(14) (15)
= 55(0. M(:.i) of M holds a sequence m 1 .k). M=zeros(k.19375) + 330(0.

m 2 . This random convergence is analyzed in Chapter 7.large. the sequences always converges to E[X ]. . m n gets close to E[X ] = 5. that we generate is random. Although each sequence m 1 . .
14
. .

To ﬁnd c.5] = 1 − FY (1. We will evaluate this integral using integration by parts:
∞ −∞
f X (x) d x =
0
∞
cxe−x/2 d x
∞ 0
(1)
∞ 0
= −2cxe−x/2
=0
+
2ce−x/2 d x
(2)
= −4ce−x/2
∞ 0
= 4c
(3)
Thus c = 1/4 and X has the Erlang (n = 2. we use ∞ the fact that −∞ f X (x) d x = 1.2 (1) First we will ﬁnd the constant c and then we will sketch the PDF. λ = 1/2) PDF
0.5) = 1 − (1.2 0.1 0 0 5 x 10 15
f X (x) =
(x/4)e−x/2 x ≥ 0 0 otherwise
fX(x)
(4)
15
.5)/4 = 5/8 Quiz 3.Quiz Solutions – Chapter 3
Quiz 3.5 0 0 2 y 4
⎧ y<0 ⎨ 0 y/4 0 ≤ y ≤ 4 FY (y) = ⎩ 1 y>4
(1)
From the CDF FY (y).1 The CDF of Y is
1 FY(y) 0. we can calculate the probabilities: (1) P[Y ≤ −1] = FY (−1) = 0 (2) P[Y ≤ 1] = FY (1) = 1/4 (3) P[2 < Y ≤ 3] = FY (3) − FY (2) = 3/4 − 2/4 = 1/4 (4) P[Y > 1.5] = 1 − P[Y ≤ 1.

we ﬁrst note X is a nonnegative random variable so that FX (x) = 0 for all x < 0. P [0 ≤ X ≤ 4] = FX (4) − FX (0) = 1 − 3e−2 .
(1)
(1) The expected value of Y is E [Y ] =
∞ −∞
y f Y (y) dy =
1 −1
(3/2)y 3 dy = (3/8)y 4
1 −1
= 0. (2) The second moment of Y is E Y2 =
∞ −∞
y 2 f Y (y) dy =
1 −1
(3/2)y 4 dy = (3/10)y 5
1 −1
= 3/5.
(3)
16
. For x ≥ 0. P [−2 ≤ X ≤ 2] = FX (2) − FX (−2) = 1 − 3e−1 ..e. (9)
(10)
Quiz 3. f Y (y) = f Y (−y)).5 0 0 5 x 10 15
FX (x) =
1− 0
x 2
+ 1 e−x/2 x ≥ 0 otherwise
(8)
(3) From the CDF FX (x). (4) Similarly.3 The PDF of Y is
3 fY(y) 2 1 0 −2 0 y 2
f Y (y) =
3y 2 /2 −1 ≤ y ≤ 1. FX (x) =
0 x
f X (y) dy =
0
x
y −y/2 e dy 4
(5) (6) (7)
x x 1 y − e−y/2 dy = − e−y/2 − 2 2 0 0 x −x/2 =1− e − e−x/2 2
The complete expression for the CDF is
1 FX(x) 0. 0 otherwise.(2) To ﬁnd the CDF FX (x).
(2)
Note that the above calculation wasn’t really necessary because E[Y ] = 0 whenever the PDF f Y (y) is an even function (i.

2 fX(x) 0 −5 x ← fX(x) ← f (y) Y 0 y 5
17
.
(5)
(z) function and Table 3.
(3)
(4)
The complete expression for the PDF of X is √ √ √ 1/(6 3) 3 − 3 3 ≤ x < 3 + 3 3.5 Each of the requested probabilities can be calculated using or Q(z) and Table 3. f X (x) = 0 otherwise.6 to write E [X ] = This implies a + b = 6. a+b =3 2 Var[X ] = (b − a)2 = 9. Since E[X ] = 3 and Var[X ] = 9. √ b = 3 + 3 3. (4) The standard deviation of Y is σY = Quiz 3. (1) √ Var[Y ] = √ 3/5.
fY(y) 0.4 (1) When X is an exponential (λ) random variable. The only valid solution with a < b is √ a = 3 − 3 3. Quiz 3. To ﬁnd a and b. we apply Theorem 3. (4)
(2) We know X is a uniform (a. it is important to remember that as the standard deviation increases. 0 otherwise. The PDF of X is f X (x) = (1/3)e−x/3 x ≥ 0. the peak value of the Gaussian PDF goes down. We start with the sketches.2. However. we must have λ = 1/3. b) random variable.(3) The variance of Y is Var[Y ] = E Y 2 − (E [Y ])2 = 3/5.4 0.1
(1) The PDFs of X and Y are shown below. E[X ] = 1/λ and Var[X ] = 1/λ2 . The fact that Y has twice the standard deviation of X is reﬂected in the greater spread of f Y (y). 12 (2)
√ b − a = ±6 3.

5 0 −2
(1.5] = Q(3. ⎩ 1 x ≥ 1. (3) P[X = 1] = FX (1+ ) − FX (1− ) = 1 − 1/2 = 1/2.(2) Since X is Gaussian (0.5] = Q( 3. ⎨ 0 FX (x) = (x + 1)/4 −1 ≤ x < 1. (4) We ﬁnd the PDF f Y (y) by taking the derivative of FY (y). (3) Since Y is Gaussian (0. P [−1 < Y ≤ 1] = FY (1) − FY (−1) 1 −1 = − σY σY (3) =2 1 − 1 = 0.6826.5 ) = Q(1. 2).0401. 2).5) = 2.75) = 1 − 2 0.33 × 10−4 . since X is Gaussian (0.7
18
.
(1)
The following probabilities can be read directly from the CDF: (1) P[X ≤ 1] = FX (1) = 1. 1). (2) P[X < 1] = FX (1− ) = 1/2. (5) Since Y is Gaussian (0. P[X > 3. P[Y > 3. Quiz 3. ⎨ 1/4 f X (x) = (1/2)δ(x − 1) x = 1.5 fX(x)
0.6 The CDF of X is
1 FX(x) 0.75) =
0 x
2
⎧ x < −1.383. 2 (4) (1) (2)
(4) Again. ⎩ 0 otherwise. 1). P [−1 < X ≤ 1] = FX (1) − FX (−1) = (1) − (−1) = 2 (1) − 1 = 0.5
0 −2
0 x
2
⎧ −1 ≤ x < 1.
(2)
Quiz 3. The resulting PDF is
0.

(3)
(3) Since X is nonnegative. (4) By taking the derivative of FY (y). Lastly. FY (y) = y−y ⎩ 1 y ≥ 1. Also. because Y ≤ 1. Note that when y < 0 or y > 1. FX (x) = 0 for x < 0.
1.6 . Y is also nonnegative.
(5)
0.5 0 −1
Y
(4)
0
As expected. FX (x) =
x −∞
f X (y) dy =
0
x
(1 − y/2) dy = x − x 2 /4. Thus FY (y) = 0 for y < 0. Using the CDF FX (x).5 0 −1 0 1 y 2 3
1 y
2
3
⎧ y < 0.8 (1) P[Y ≤ 6] =
6 −∞ f Y (y) dy
=
6 0 (1/10) dy
= 0. FX (x) = 1 for x ≥ 2 since its always true that x ≤ 2.5 0 −1
X
0
1 x
2
3
⎧ x < 0.(1) Since X is always nonnegative. for 0 < y < 1.25
f Y (y) =
1 − y/2 + (1/4)δ(y − 1) 0 ≤ y ≤ 1 0 otherwise
Y
(6)
Quiz 3.
19
.
(1)
The complete CDF of X is
1 F (x) 0. the complete expression for the CDF of Y is
1 F (y) 0. the PDF is zero. we see that the jump in FY (y) at y = 1 is exactly equal to P[Y = 1]. FY (y) = 1 for all y ≥ 1. Finally.5 f (y) 1 0. FX (x) = x−x ⎩ 1 x > 2. FY (y) = P [Y ≤ y] = P [X ≤ y] = FX (y) . for 0 ≤ x ≤ 2. we obtain the PDF f Y (y). ⎨ 0 2 /4 0 ≤ x ≤ 2. ⎨ 0 2 /4 0 ≤ y < 1. Also.
(2)
(2) The probability that Y = 1 is P [Y = 1] = P [X ≥ 1] = 1 − FX (1) = 1 − 3/4 = 1/4.

20
.15.(2) From Deﬁnition 3.
(3)
(5) From the conditional PDF f Y |Y ≤6 (y). 0 otherwise. then T = T + 2 has PDF f T (t) = f T |T >2 (t).2 .9 A natural way to produce random variables with PDF f T |T >2 (t) is to generate samples of T with PDF f T (t) and then to discard those samples which fail to satisfy the condition T > 2.1). 10
(2)
(4) From Deﬁnition 3.m)
generates the vector t. if (x>2) t(i+1)=x. t=zeros(m.
(1)
1 dy = 0. = otherwise. we can calculate the conditional expectation E [Y |Y > 8] =
∞ −∞
y f Y |Y >8 (y) dy =
10 8
y dy = 9.1).0+exponentialrv(1/3. the conditional PDF of Y given Y > 8 is f Y |Y >8 (y) =
f Y (y) P[Y >8]
0
y > 8.15. the conditional PDF of Y given Y ≤ 6 is f Y |Y ≤6 (y) = (3) The probability Y > 8 is P [Y > 8] =
8 10 f Y (y) P[Y ≤6]
0
y ≤ 6.lambda=1/3. while (i<m). 2
(5)
Quiz 3. In this case the command
t=2. x=exponentialrv(lambda. we can calculate the conditional expectation E [Y |Y ≤ 6] =
∞ −∞
y f Y |Y ≤6 (y) dy =
6 0
y dy = 3.
1/6 0 ≤ y ≤ 6. 6
(4)
(6) From the conditional PDF f Y |Y >8 (y). Here is a M ATLAB function that uses this method:
function t=t2rv(m) i=0. end end
A second method exploits the fact that if T is an exponential (λ) random variable.
1/2 8 < y ≤ 10. = otherwise. i=i+1. 0 otherwise.

(2) FX.G (1. This result is given in Theorem 4. 0) + PQ. 1) = 0.6 (2) The probability that Q = G is P [Q = G] = PQ.12 + 0.Y (−∞.G (0.08 = 0.18 + 0.G (0.24 + 0.16 + 0.18 (3) The probability that G > 1 is
3 1
(1) (2)
(3)
P [G > 1] =
g=2 q=0
PQ. (3) FX.16 + 0.Y (∞.G (0.G (q. (1) FX.G (0. y) = P[X ≤ ∞. 1) + PQ.Quiz Solutions – Chapter 4
Quiz 4. 2) = P[X ≤ −∞. 0) + PQ.Y (∞.24 + 0. Y ≤ ∞] = 1. ∞) = P[X ≤ ∞. g)
(4) (5)
= 0.1. Quiz 4.78
21
. −∞) = P[X ≤ ∞. (4) FX. 3) = 0. we can calculate the requested probabilities by summing the PMF over those values of Q and G that correspond to the event.06 + 0. g)
(6) (7)
= 0. Y ≤ y] = P[Y ≤ y] = FY (y).G (q.12 = 0. (1) The probability that Q = 0 is P [Q = 0] = PQ. Y ≤ −∞] = 0 since Y cannot take on the value −∞.G (0.Y (∞.6 (4) The probability that G > Q is
1 3
P [G > Q] =
q=0 g=q+1
PQ.2 From the joint PMF of Q and G given in the table.18 + 0. Y ≤ 2] ≤ P[X ≤ −∞] = 0 since X cannot take on the value −∞.24 + 0. 2) + PQ.08 = 0.1 Each value of the joint CDF can be found by considering the corresponding probability.12 + 0.

y) d x d y = =c
cx y d x dy y
0 2 0
(1) dy
2 0
x 2 /2
1 0
(2) =c (3)
= (c/2) Thus c = 1. b)
(1)
For each value of h. the marginal PMF of B is
1
PB (b) =
h=−1
PH.2 0.Y (x. Speciﬁcally.B (h. b)
(2)
For each value of b.Quiz 4.1 0 0.2.B (h.1 0. we write P [A] =
A
y dy = (c/4)y 2
f X. The easiest way to calculate these marginal PMFs is to simply sum each row and column: PH.4 To ﬁnd the constant c.2 h=0 h=1 0.3. yielding
2 1
Y
P [A] =
0
π/2 0 1 0
1
r 2 sin θ cos θ r dr dθ
π/2 0 2 π/2
(5) (6)
A 1 X
= =
r 3 dr ⎛
1 0
sin θ cos θ dθ ⎞ ⎠ = 1/8
r 4 /4
⎝ sin θ 2
(7)
0
22
. b) b = 0 b = 2 b = 4 PH (h) h = −1 0 0.
2 0 0 2 1
f X. y) d x d y
(4)
To integrate over A.B (h. we apply
∞ ∞ −∞ −∞ ∞ ∞ −∞ −∞
(3)
f X. the marginal PMF of H is PH (h) =
b=0.2 PB (b) 0.5 0.Y (x.3 Quiz 4.1 0.4 0. y = r sin θ and d x d y = r dr dθ .2 0.Y (x.1 0 0. this corresponds to calculating the row sum across the table of the joint PMF. To calculate P[A].4
PH. Similarly. we convert to polar coordinates using the substitutions x = r cos θ .6 0.3 By Theorem 4. y) d x d y = 1. this corresponds to the column sum down the table of the joint PMF.

000 l = 7.2 t = 270 ⎪ ⎪ ⎪ ⎪ 0. b) l = 518.20 (T =36) 0.Y (x.B (l.05 (T =180) 0.10 (T =360) b = 28. We can write these down on the table for the joint PMF of L and B as follows: PL . the complete expression for the PDF of Y is f Y (y) = Quiz 4.05 t = 18 ⎪ ⎪ ⎪ 0. writing down the PMF of T is straightforward. 592. 90 ⎪ ⎪ ⎨ 0. 400 l = 2.8. For each pair of values of L and B. For 0 ≤ y ≤ 1.5 By Theorem 4. 800 0.1 t = 120 PT (t) = ⎪ 0. we can calculate the time T needed for the transfer.20 (T =270) (3 + 6y 2 )/5 0 ≤ y ≤ 1 0 otherwise (6)
From the table. ⎧ ⎪ 0. y) dy
(1)
For x < 0 or x > 1.10 (T =120) 0.1 t = 360 ⎪ ⎪ ⎩ 0 otherwise 23
(1)
. f X (x) = 0.10 (T =24) 0.05 t = 180 ⎪ ⎪ ⎪ 0. 776.20 (T =90) 0. f X (x) = 6 5
1 0
(x + y 2 ) dy =
6 x y + y 3 /3 5
y=1 y=0
6x + 2 6 = (x + 1/3) = 5 5
(2)
The complete expression for the PDf of X is f X (x) = (6x + 2)/5 0 ≤ x ≤ 1 0 otherwise (3)
By the same method we obtain the marginal PDF for Y . f Y (y) = =
∞ −∞ 6 1
f X. 600 0. the marginal PDF of X is f X (x) =
∞ −∞
f X. y) dy (x + y 2 ) d x = 6 2 x /2 + x y 2 5
x=1 x=0
(4) 6 3 + 6y 2 = (1/2 + y 2 ) = 5 5 (5)
5
0
Since f Y (y) = 0 for y < 0 or y > 1.6 (A) The time required for the transfer is T = L/B. 000 b = 14. For 0 ≤ x ≤ 1.2 t = 36.00 (T =540) b = 21.Y (x.Quiz 4. 400 0.1 t = 24 ⎪ ⎪ ⎪ ⎪ 0.05 (T =18) 0.

15 0.3 0.4 PL (l) 0.25
(7)
(8)
(1)
(2)
(3)
. Thus f W (0) = 0 and f W (1) = 1.5. integrating over the region W ≤ w is fairly complex.5.T (l. For 0 < w < 1. t) l=1 l=2 l=3 PT (t) (1) The expected value of L is E [L] = 1(0.25) = 2.5 0.25) = 4.5) + 3(0. we observe that since 0 ≤ X ≤ 1 and 0 ≤ Y ≤ 1.7 (A) It is helpful to ﬁrst make a table that includes the marginal PMFs.25) + 22 (0. PL . the variance of L is Var [L] = E L 2 − (E [L])2 = 0.5) + 32 (0.(B) First. As shown below.15 0.6 t = 60 0. Since the second moment of L is E L 2 = 12 (0.
Y 1 w w 1 XY > w
FW (w) = 1 − P [X Y > w] =1− =1−
1 1 w w/x 1 w
(2) (3) (4) (5) (6)
dy dx
XY = w X
(1 − w/x) d x
= 1 − x − w ln x|x=1 x=w
= 1 − (1 − w + w ln w) = w − w ln w The complete expression for the CDF is ⎧ w<0 ⎨ 0 FW (w) = w − w ln w 0 ≤ w ≤ 1 ⎩ 1 w>1 By taking the derivative of the CDF. we calculate the CDF FW (w) = P[W ≤ w]. 24 t = 40 0. W = X Y satisﬁes 0 ≤ W ≤ 1.2 0. The calculus is simpler if we integrate over the region X Y > w.1 0. Speciﬁcally.25 0.25) + 2(0.1 0. we ﬁnd the PDF is ⎧ 0 w<0 d FW (w) ⎨ f W (w) = = − ln w 0 ≤ w ≤ 1 ⎩ dw 0 w>1 Quiz 4.

T = 0.15) + 1(60)(0. the covariance of L and T is Cov [L .2) + 3(60)(0.15) + 2(40)(0.16(a).
(11)
(B) As in the discrete case.3) + 3(40)(0.4) = 48. for 0 ≤ y ≤ 2. the correlation coefﬁcient is ρ L .60 l=1
lt PL T (lt)
(7) (8) (9) (10)
= 1(40)(0. For 0 ≤ x ≤ 1.Y (x. T ] = 0.6) + 60(0. f Y (y) =
∞ −∞
f X. (3) The correlation is
3
(4)
(5)
(6)
E [L T ] =
t=40. The second moment of T is E T 2 = 402 (0.(2) The expected value of T is E [T ] = 40(0. y) d x =
0
2
xy dx =
1 2 x y 2
x=1
=
x=0
y 2
(13)
The complete expressions for the marginal PDFs are f X (x) = 2x 0 ≤ x ≤ 1 0 otherwise f Y (y) = y/2 0 ≤ y ≤ 2 0 otherwise (14)
From the marginal PDFs. T ] = E [L T ] − E [L] E [T ] = 96 − 2(48) = 0 (5) Since Cov[L . it is straightforward to calculate the various expectations.
25
. Thus Var[T ] = E T 2 − (E [T ])2 = 2400 − 482 = 96. the calculations become easier if we ﬁrst calculate the marginal PDFs f X (x) and f Y (y).1) = 96 (4) From Theorem 4.Y (x. y) dy =
0
2
1 x y dy = x y 2 2
y=2
= 2x
y=0
(12)
Similarly.1) + 2(60)(0.4) = 2400.6) + 602 (0. f X (x) =
∞ −∞
f X.

T (3.T |A (l.Y = 0. T ) = (2. PL .T (l. (2) The ﬁrst and second moments of Y are E [Y ] = E Y2 4 1 2 y dy = 3 −∞ 0 2 ∞ 2 1 = y 2 f Y (y) dy = y 3 dy = 2 −∞ 0 2 y f Y (y) dy =
∞ 2
(18) (19)
The variance of Y is Var[Y ] = E[Y 2 ] − (E[Y ])2 = 2 − 16/9 = 2/9. Y ] = E [X Y ] − E [X ] E [Y ] = 2 8 − 9 3 4 3 = 0.9. (22)
(5) Since Cov[X. P [A] = P [V > 80] = PL .t) P[A]
(1)
0
lt > 80 otherwise
(2)
.(1) The ﬁrst and second moments of X are E [X ] = E X2 =
∞ −∞ ∞ −∞
x f X (x) d x =
0
1
2x 2 d x =
1
2 3 1 2
(15) (16) (17)
x 2 f X (x) d x =
0
2x 3 d x =
The variance of X is Var[X ] = E[X 2 ] − (E[X ])2 = 1/18. 60) + PL . Quiz 4. dy = 3
y3 3
=
0
8 9
(21)
(4) The covariance of X and Y is Cov [X.8 (A) Since the event V > 80 occurs only for the pairs (L .45 By Deﬁnition 4. Y ] = 0.T (3. 40) + PL .T (2. (L . T ) = (3. 60). T ) = (3. (3) The correlation of X and Y is E [X Y ] = =
∞ ∞
−∞ −∞ 1 2 2 2 0 0
x y f X. dy
1 0
(20)
2
x3 x y d x. the correlation coefﬁcient is ρ X. y) d x. 40) and (L . 60) = 0.Y (x. t) = 26
PL . 60).

y) /P [B] (x. t)
(5) (6)
4 1 2 = (2 · 60)2 + (3 · 40)2 + (3 · 60)2 = 18. 80/y ≤ x ≤ 3 0 otherwise 27 (12) (13)
. we ﬁrst calculate the probability of the conditioning event.Y (x.T |A (l. y) = = f X. 400 9 3 9 It follows that Var [V |A] = E V 2 |A − (E [V |A])2 = 622 2 9
(7)
(B) For continuous random variables X and Y . P [B] =
B
f X. y) d x d y = = =
60 40 60 40 60
3 80/y
xy dx dy 4000 x2 2
3
(8) dy (9) (10) (11)
y 4000
80/y
9 3200 y − 2 y 40 4000 2 9 4 3 = − ln ≈ 0.801 8 5 2
dy
The conditional PDF of X and Y is f X. t) t = 40 t = 60 l=1 0 0 l=2 0 4/9 1/3 2/9 l=3 The conditional expectation of V can be found from the conditional PMF. E [V |A] =
l t
lt PL . t)
(3) (4)
1 2 1 4 = (2 · 60) + (3 · 40) + (3 · 60) = 133 9 3 9 3
For the conditional variance Var[V |A].T |A (l.We can represent this conditional PMF in the following table: PL .Y (x. y) ∈ B 0 otherwise K x y 40 ≤ y ≤ 60.Y |B (x. we ﬁrst ﬁnd the conditional second moment E V 2 |A =
l t
(lt)2 PL .T |A (l.

**where K = (4000P[B])−1 . The conditional expectation of W given event B is E [W |B] = =
**

∞ ∞ −∞ −∞ 60 3 40

x y f X,Y |B (x, y) d x d y K x 2 y2 d x d y y2 x 3

x=3 x=80/y

(14) (15)

= (K /3) = (K /3)

80/y 60 40 60 40

dy

(16) (17) (18)

27y 2 − 803 /y dy

60 40

**= (K /3) 9y 3 − 803 ln y The conditional second moment of K given B is E W 2 |B = =
**

∞ ∞

≈ 120.78

−∞ −∞ 60 3 40

(x y)2 f X,Y |B (x, y) d x d y K x 3 y3 d x d y y3 x 4

x=3 x=80/y

(19) (20)

= (K /4)

80/y 60 40 60 40

dy

(21) (22) ≈ 16, 116.10 (23)

= (K /4)

81y 3 − 804 /y dy

60 40

= (K /4) (81/4)y 4 − 804 ln y It follows that the conditional variance of W given B is

Var [W |B] = E W 2 |B − (E [W |B])2 ≈ 1528.30 Quiz 4.9

(24)

(A) (1) The joint PMF of A and B can be found from the marginal and conditional PMFs via PA,B (a, b) = PB|A (b|a)PA (a). Incorporating the information from the given conditional PMFs can be confusing, however. Consequently, we can note that A has range S A = {0, 2} and B has range S B = {0, 1}. A table of the joint PMF will include all four possible combinations of A and B. The general form of the table is PA,B (a, b) b=0 b=1 a=0 PB|A (0|0)PA (0) PB|A (1|0)PA (0) PB|A (0|2)PA (2) PB|A (1|2)PA (2) a=2 28

Substituting values from PB|A (b|a) and PA (a), we have b=0 b=1 PA,B (a, b) a=0 (0.8)(0.4) (0.2)(0.4) (0.5)(0.6) (0.5)(0.6) a=2 or PA,B (a, b) b = 0 b = 1 a=0 0.32 0.08 0.3 0.3 a=2

**(2) Given the conditional PMF PB|A (b|2), it is easy to calculate the conditional expectation
**

1

E [B|A = 2] =

b=0

b PB|A (b|2) = (0)(0.5) + (1)(0.5) = 0.5

(1)

(3) From the joint PMF PA,B (a, b), we can calculate the the conditional PMF ⎧ 0.32/0.62 a = 0 PA,B (a, 0) ⎨ PA|B (a|0) = = 0.3/0.62 a = 2 (2) ⎩ PB (0) 0 otherwise ⎧ ⎨ 16/31 a = 0 = 15/31 a = 2 (3) ⎩ 0 otherwise (4) We can calculate the conditional variance Var[A|B = 0] using the conditional PMF PA|B (a|0). First we calculate the conditional expected value E [A|B = 0] =

a

a PA|B (a|0) = 0(16/31) + 2(15/31) = 30/31

(4)

**The conditional second moment is E A2 |B = 0 =
**

a

a 2 PA|B (a|0) = 02 (16/31) + 22 (15/31) = 60/31 (5)

The conditional variance is then Var[A|B = 0] = E A2 |B = 0 − (E [A|B = 0])2 = (B) (1) The joint PDF of X and Y is f X,Y (x, y) = f Y |X (y|x) f X (x) = (2) From the given conditional PDF f Y |X (y|x), f Y |X (y|1/2) = 29 8y 0 ≤ y ≤ 1/2 0 otherwise (8) 6y 0 ≤ y ≤ x, 0 ≤ x ≤ 1 0 otherwise (7) 960 961 (6)

**(3) The conditional PDF of Y given X = 1/2 is f X |Y (x|1/2) = f X,Y (x, 1/2)/ f Y (1/2). To ﬁnd f Y (1/2), we integrate the joint PDF. f Y (1/2) = Thus, for 1/2 ≤ x ≤ 1, f X |Y (x|1/2) = f X,Y (x, 1/2) 6(1/2) =2 = f Y (1/2) 3/2 (10)
**

∞ −∞

f X,1/2 ( ) d x =

1 1/2

6(1/2) d x = 3/2

(9)

(4) From the pervious part, we see that given Y = 1/2, the conditional PDF of X is uniform (1/2, 1). Thus, by the deﬁnition of the uniform (a, b) PDF, Var [X |Y = 1/2] = Quiz 4.10 (A) (1) For random variables X and Y from Example 4.1, we observe that PY (1) = 0.09 and PX (0) = 0.01. However, PX,Y (0, 1) = 0 = PX (0) PY (1) (1) (1 − 1/2)2 1 = 12 48 (11)

Since we have found a pair x, y such that PX,Y (x, y) = PX (x)PY (y), we can conclude that X and Y are dependent. Note that whenever PX,Y (x, y) = 0, independence requires that either PX (x) = 0 or PY (y) = 0. (2) For random variables Q and G from Quiz 4.2, it is not obvious whether they are independent. Unlike X and Y in part (a), there are no obvious pairs q, g that fail the independence requirement. In this case, we calculate the marginal PMFs from the table of the joint PMF PQ,G (q, g) in Quiz 4.2. PQ,G (q, g) g = 0 g = 1 g = 2 g = 3 PQ (q) q=0 0.06 0.18 0.24 0.12 0.60 0.04 0.12 0.16 0.08 0.40 q=1 PG (g) 0.10 0.30 0.40 0.20 Careful study of the table will verify that PQ,G (q, g) = PQ (q)PG (g) for every pair q, g. Hence Q and G are independent. (B) (1) Since X 1 and X 2 are independent, f X 1 ,X 2 (x1 , x2 ) = f X 1 (x1 ) f X 2 (x2 ) = (1 − x1 /2)(1 − x2 /2) 0 ≤ x1 ≤ 2, 0 ≤ x2 ≤ 2 0 otherwise 30 (2) (3)

(2) Let FX (x) denote the CDF of both X 1 and X 2 . The CDF of Z = max(X 1 , X 2 ) is found by observing that Z ≤ z iff X 1 ≤ z and X 2 ≤ z. That is, P [Z ≤ z] = P [X 1 ≤ z, X 2 ≤ z] = P [X 1 ≤ z] P [X 2 ≤ z] = [FX (z)]2 (4) (5)

To complete the problem, we need to ﬁnd the CDF of each X i . From the PDF f X (x), the CDF is ⎧ x <0 ⎨ 0 x 2 /4 0 ≤ x ≤ 2 FX (x) = f X (y) dy = (6) x−x ⎩ −∞ 1 x >2 Thus for 0 ≤ z ≤ 2, FZ (z) = (z − z 2 /4)2 (7)

The complete expression for the CDF of Z is ⎧ z<0 ⎨ 0 2 /4)2 0 ≤ z ≤ 2 FZ (z) = (z − z ⎩ 1 z>1

(8)

Quiz 4.11 This problem just requires identifying the various terms in Deﬁnition 4.17 and Theorem 4.29. Speciﬁcally, from the problem statement, we know that ρ = 1/2, µ1 = µ X = 0, and that σ1 = σ X = 1, σ2 = σY = 1. (2) (1) Applying these facts to Deﬁnition 4.17, we have 1 2 2 e−2(x −x y+y )/3 . f X,Y (x, y) = √ 3π 2 (3) µ2 = µY = 0, (1)

(2) By Theorem 4.30, the conditional expected value and standard deviation of X given Y = y are 2 E [X |Y = y] = y/2 σ X = σ1 (1 − ρ 2 ) = 3/4. ˜ (4) When Y = y = 2, we see that E[X |Y = 2] = 1 and Var[X |Y = 2] = 3/4. The conditional PDF of X given Y = 2 is simply the Gaussian PDF 1 2 e−2(x−1) /3 . f X |Y (x|2) = √ 3π/2 (5)

31

. 0 otherwise. x) PMF. given X = x. 3. Instead.1).y’].px. PY |X (y|x) = 1/x y = 1. px=0.4].Quiz 4.
32
. 4. First we observe that X has the discrete uniform (1. PX (x) = 1/4 x = 1. This observation prompts the following program:
function xy=dtrianglerv(m) sx=[1. and an independent uniform (0. y=ceil(x.1)). 1) random variable U .28. Y has a discrete uniform (1. x=finiterv(sx. 2. .2.25*ones(4.12 One straightforward method is to follow the approach of Example 4.3.m). x 0 otherwise (1)
Given X = x.*rand(m. we use an alternate approach. . . we can generate a sample value of Y with a discrete uniform (1. That is. x) PMF via Y = xU . xy=[x’. Also. 4) PMF.

P [C] =
0 1/2 y2 1/2 y4
dy2
0 1/2
dy1
0
dy4
0 1/2
4dy3 = 1/4. 6 d x1 = 6x2 . X 2 = y2 + y1 . 6 d x2 = 6(x3 − x1 ). Y1 = X 1 . .X 3 (x2 .
(1) (2) (3)
x2 x2 0 x3 x1
In particular.3 First we note that each marginal PDF is nonzero only if any subset of the xi obeys the ordering contraints 0 ≤ x 1 ≤ x2 ≤ x3 ≤ 1. X 3 − X 2 = y3 ] = P [X 1 = y1 . X 3 = y3 + y2 + y1 ] = (1 − p)3 p y1 +y2 +y3 (1) (2) (3) (4)
By deﬁning the vector a = 1 1 1 . we must keep in mind that f X 1 .
(1) (2)
=4
0
y2 dy2
0
y4 dy4
Quiz 5.}. and that f X 1 . y3 ∈ {1. x2 ) = 0 unless 0 ≤ x 1 ≤ x2 ≤ 1. . Speciﬁcally. PY (y) = P [Y1 = y1 . . . x3 ) = 0 unless 0 ≤ x 1 ≤
33
. Since 0 < X 1 < X 2 < X 3 . 2.Quiz Solutions – Chapter 5
Quiz 5. each Yi must be a strictly positive integer. Thus.X 3 (x2 . Y3 = y3 ] = P [X 1 = y1 .} 0 otherwise (5)
Quiz 5. X 2 − X 1 = y2 . Y2 = X 2 − X 1 and Y3 = X 3 − X 2 . y2 . f X 2 .X 3 (x1 .X 2 (x1 . x3 ) =
∞ −∞ ∞ −∞ ∞ −∞
f X (x) d x3 = f X (x) d x1 = f X (x) d x2 =
1
6 d x3 = 6(1 − x2 ).2 By deﬁnition of A. 2. . we have f X 1 .1 We ﬁnd P[C] by integrating the joint PDF over the region of interest. x3 ) = 0 unless 0 ≤ x2 ≤ x3 ≤ 1. the complete expression for the joint PMF of Y is PY (y) = (1 − p) p a y y1 . x2 ) = f X 2 .X 3 (x1 . . Y2 = y2 . for y1 . y2 .X 2 (x1 . Within these constraints. y3 ∈ {1. x3 ) = f X 1 .

Y2 W= Y3 . x3 ) = 6(1 − x2 ) 0 ≤ x1 ≤ x2 ≤ 1 0 otherwise 6x2 0 ≤ x2 ≤ x3 ≤ 1 0 otherwise 6(x3 − x1 ) 0 ≤ x1 ≤ x3 ≤ 1 0 otherwise (4) (5) (6)
Now we can ﬁnd the marginal PDFs. w) = 4 0 ≤ v1 ≤ v2 ≤ 1. When 0 ≤ xi ≤ 1 for each xi . x2 ) = f X 2 . the components have dependencies as a result of the ordering constraints Y1 ≤ Y2 and Y3 ≤ Y4 . f X 1 (x1 ) = f X 2 (x2 ) = f X 3 (x3 ) =
∞ −∞ ∞ −∞ ∞ −∞
f X 1 . The complete expressions are f X 1 . x3 ) d x3 = f X 2 .W (v.X 2 (x1 .X 3 (x1 .X 2 (x1 . We can separate these constraints by creating the vectors V= The joint PDF of V and W is f V. Y4 (1)
34
.X 3 (x2 .X 3 (x2 .x3 ≤ 1.X 3 (x2 . 0 ≤ w1 ≤ w2 ≤ 1 0 otherwise (2) Y1 . x3 ) = f X 1 . x2 ) d x2 = f X 2 . x3 ) d x2 =
1 x1 1
6(1 − x2 ) d x2 = 3(1 − x1 )2 6x2 d x3 = 6x2 (1 − x2 )
2 6x2 d x2 = 3x3
(7) (8) (9)
x2 x3 0
The complete expressions are f X 1 (x1 ) = f X 2 (x2 ) = f X 3 (x3 ) = 3(1 − x1 )2 0 ≤ x1 ≤ 1 0 otherwise 6x2 (1 − x2 ) 0 ≤ x2 ≤ 1 0 otherwise
2 3x3 0 ≤ x3 ≤ 1 0 otherwise
(10) (11) (12)
Quiz 5.4 In the PDF f Y (y).

for p1 = 0. 5} ⎩ 0 otherwise We can ﬁnd the marginal PMF for each X i from the joint PMF PX (x). . That is. 0 otherwise f W (w) = 2 0 ≤ w 1 ≤ w2 ≤ 1 0 otherwise (8)
It is easy to verify that f V. .5 (A) Referring to Theorem 1.3) random variable.W (v. f V (v) = =
0 1
f V. A and R. conﬁrming that V and W are independent vectors. In ﬁve trials. 5 0 otherwise 35
5 x
(2)
. .1. . the vector X = X 1 X 2 X 3 indicating the number of outcomes of each subexperiment has the multinomial PMF ⎧ 5 ⎨ x1 .W (v. 0.3. PX i (x) = pix (1 − pi )5−x x = 0. PX (x) = (1) x1 . For 0 ≤ v1 ≤ v2 ≤ 1.3)x1 (0. w) dv1 dv2
1 0 1 v1
(6) (7)
4 dv2
dv1 = 2
It follows that V and W have PDFs f V (v) = 2 0 ≤ v1 ≤ v2 ≤ 1 . .6 and p3 = 0. p2 = 0.19. each test is a subexperiment with three possible outcomes: L. . 1. If we view each test as a trial with success probability P[L] = 0. 1. w) dw1 dw2
1 w1 1 0
(3) (4) (5)
4 dw2
dw1
= Similarly. 0. however it is simpler to just start from ﬁrst principles and observe that X 1 is the number of occurrences of L in ﬁve independent tests.x2 . f W (w) = =
4(1 − w1 ) dw1 = 2
f V. . for 0 ≤ w1 ≤ w2 ≤ 1. we see that X 1 is a binomial (n.3. 0.6)x2 (0.1)x3 x1 + x2 + x3 = 5.1) random variable.6) random variable and X 3 is a binomial (5. p) = (5.W (v. X 2 is a binomial (5. x3 ∈ {0. Similarly. w) = f V (v) f W (w). .We must verify that V and W are independent. x2 . Quiz 5.x3 (0.

we must use Theorem 5.6 We start by ﬁnding the components E[X i ] = the marginal PDFs f X i (x) found in Quiz 5.
(1) (2) (3)
E [X 2 ] =
0
1
E [X 3 ] =
0
1
To ﬁnd the correlation matrix R X . Quiz 5.1)] 2!2!1! = 0.288 PW (5) = PX 1 (5) + PX 2 (5) + PX 3 (5) = 0. In particular. . 3x 3 d x = 3/4. since X 1 + X 2 + X 3 = 5 and since each X i is non-negative. 1) 5![0. for w = 3. we use
3x(1 − x)2 d x = 1/4.From the marginal PMFs.1458 = (3) (4) (5)
In addition.6)2 (0. PW (0) = PW (1) = 0. To do so. f 3 X 2 2 2 2 (1/8)e−(y3 −4)/2 4 ≤ y1 ≤ y2 ≤ y3 = 0 otherwise (9) (10) (6) (7) (8)
Note that for other matrices A.0802 (B) Since each Yi = 2X i + 4.1)2 + 0. and w = 5. Hence. or X 3 = w occurs. PW (2) = PX (1. the event W = w occurs if and only if one of the mutually exclusive events X 1 = w. 2) + PX (2.10 to write f Y (y) = y1 − 4 y2 − 4 y3 − 4 1 .3(0. X 2 = w. we can apply Theorem 5. Thus. 6x 2 (1 − x) d x = 1/2.3: E [X 1 ] =
0 1
∞ −∞
x f X i (x) d x of µ X . Furthermore.486 PW (4) = PX 1 (4) + PX 2 (4) + PX 3 (4) = 0.6)(0.6 to ﬁnd the PMF of W .6)2 (0. we see that X 1 . We start with
36
.32 (0. 2. we need to ﬁnd E[X i X j ] for all i and j. 2. 1. 2) + PX (2. X 2 and X 3 are not independent.32 (0. w = 4. the constraints on y resulting from the constraints 0 ≤ X 1 ≤ X 2 ≤ X 3 can be much more complicated.1)2 + 0. PW (3) = PX 1 (3) + PX 2 (3) + PX 3 (3) = 0.

x3 =1 x3 =x1
=
0
1
3 2 2 (2x1 x3 − 3x1 x3 )
d x1
=
0
1
2 4 [2x1 − 3x1 + x1 ] d x1 = 1/5.the second moments: E
2 X1
=
0
1
3x 2 (1 − x)2 d x = 1/10.
(4) (5) (6)
2 E X2 = 2 E X3 =
1 0 1 0
Using marginal PDFs from Quiz 5. 1/4 3/8 ⎦ = 80 1 2 3 3/8 9/16 (17)
(18)
.3. 6x 3 (1 − x) d x = 3/10. the cross terms are E [X 1 X 2 ] = = =
0 ∞ ∞ −∞ −∞ 1 1 0 1 x1 3 4 [x1 − 3x1 + 2x1 ] d x1 = 3/20. d x1 d x2 d x1
(7) (8) (9) (10) (11) (12) (13) (14)
6x1 x2 (1 − x2 ) d x2
E [X 2 X 3 ] =
0
1
= E [X 1 X 3 ] =
0
2 4 [3x2 − 3x2 ] d x2 = 2/5 1 x1
1
6x1 x3 (x3 − x1 ) d x3 d x1 . 1/5 2/5 3/5 Vector X has covariance matrix C X = R X − E [X] E [X] ⎡ ⎤ ⎡ ⎤ 1/10 3/20 1/5 1/4 ⎣3/20 3/10 2/5⎦ − ⎣1/2⎦ = 1/5 2/5 3/5 3/4 ⎡ ⎤ ⎡ 1/10 3/20 1/5 1/16 ⎣3/20 3/10 2/5⎦ − ⎣ 1/8 = 1/5 2/5 3/5 3/16 37
(15)
(16) 1/4 1/2 3/4 ⎤ ⎡ ⎤ 3 2 1 1/8 3/16 1 ⎣ 2 4 2⎦ . x2 ) .
Summarizing the results. 1 x2 1 0 2 6x2 x3 d x3 d x2
x1 x2 f X 1 .X 2 (x1 . X has correlation matrix ⎡ ⎤ 1/10 3/20 1/5 R X = ⎣3/20 3/10 2/5⎦ . 3x 4 d x = 3/5.

m:
>> julytemps([70 75 80 85 90 95]) ans = 0.0221 0. the ﬁrst two lines generate the 31 × 31 covariance matrix CT. just a Gaussian random variable.16 tells us that Y is a 1 dimensional Gaussian vector.0. 1 −1 b= 2 .0000
1.00002844263128 0.
function p=julytemps(T). Quiz 5. Theorem 5.9779
1.0000..99997155736872 0.(1:31)). computing the covariance matrix by calculus can be a time consuming task.0000
Note that P[T ≤ 70] is not actually zero and that P[T ≤ 90] is not actually 1.
Here is the output of julytemps. The expected value of Y is µY = µT = 80. Next we calculate Var[Y ].m. or CT . Its just that the M ATLAB’s short format output. Var[Y ] = ACT A .1)/31.5000 0.8 First.18 that µ X = b and that C X = AA = 2 1 1 −1 2 1 5 1 = . invoked with the command format short. A=ones(31.This problem shows that even for fairly simple joint PDFs. The covariance matrix of Y is 1 × 1 and is just equal to Var[Y ].50000000000000
0.7 We observe that X = AZ + b where A= 2 1 . 0 (1)
It follows from Theorem 5. i. CY=(A’)*CT*A./(1+abs(D1-D2)). CT=36. The ﬁnal step is to use the (·) function to calculate P[Y < T ]. we observe that Y = AT where A = 1/31 1/31 · · · 1/31 .e. rounds off those probabilities. p=phi((T-80)/sqrt(CY)). Here is the long format output:
>> format long >> julytemps([70 75 80 85 90 95]) ans = Columns 1 through 4 0.
In julytemps. [D1 D2]=ndgrid((1:31).02207383067604 Columns 5 through 6 0. 1 −1 1 2 (2)
Quiz 5.16. Thus. Since T is a Gaussian random vector.0000 0. by Theorem 5.99999999922010
0.97792616932396
38
.

.0. In fact. the i./(1+abs(0:30)). ⎣ . A=ones(31. c=36. We will see in Chapters 9 and 11 that Toeplitz covariance matrices are quite common.1)/31. The function julytemps2 use the toeplitz to generate the correlation matrix CT . ⎥ .. ⎥. we see that ⎡ ⎤ c0 c1 · · · c30 . .The ndgrid function is a useful to way calculate many covariance matrices. in this problem. c1 ⎦ . c30 · · · c1 c0
(2)
This covariance matrix is known as a symmetric Toeplitz matrix. CT=toeplitz(c). ⎥ ⎢ . 1 + |i − j| (1)
If we write out the elements of the covariance matrix.. M ATLAB has a toeplitz function for generating them. . However. . . C X has a special structure. j) = c|i− j| = 36 . jth element is CT (i.
39
.. ⎢ c1 c0 CT = ⎢ .
function p=julytemps2(T).. p=phi((T-80)/sqrt(CY)). CY=(A’)*CT*A.

For w > 0. f W (w) = e−3w e y
w 0
= 6 e−2w − e−3w
(3)
Since f W (w) = 0 for w < 0. K n are independent. the expected value of Wn is E [Wn ] = E [K 1 ] + · · · + E [K n ] = n E [K i ] = 2.Quiz Solutions – Chapter 6
Quiz 6. . . the random variables K 1 . otherwise. 4 0 otherwise (1)
We can write Wn in the form of Wn = K 1 + · · · + K n . .25n Quiz 6. (4)
40
. First. Var[Wn ] = Var[K 1 ] + · · · + Var[K n ] = 1. we note that the ﬁrst two moments of K i are E [K i ] = (1 + 2 + 3 + 4)/4 = 2. this integral is easy to evaluate.25 Since E[K i ] = 2. W = X + Y is nonnegative. K n denote a sequence of iid random variables each with PMF PK (k) = 1/4 k = 1. .1 Let K 1 .5 Thus the variance of K i is Var[K i ] = E K i2 − (E [K i ])2 = 7. . a conmplete expression for the PDF of W is f W (w) = 6e−2w 1 − e−w 0 w ≥ 0. the PDF of W = X + Y is f W (w) =
∞ −∞
f X (w − y) f Y (y) dy = 6
0
w
e−3(w−y) e−2y dy
(2)
Fortunately.2 Random variables X and Y have PDFs f X (x) = 3e−3x x ≥ 0 0 otherwise f Y (y) = 2e−2y y ≥ 0 0 otherwise (1) (6) (4) (2) (3)
Since X and Y are nonnegative. the variance of the sum equals the sum of the variances. .5)2 = 1. Hence.5 − (2. . . . by Theorem 6.3. By Theorem 6. .5n (5) Since the rolls are independent.5 E K i2 = (12 + 22 + 32 + 42 )/4 = 7. .5. . That is.5.

Since the expectation of the sum equals the sum of the expectations: E [W ] = α E [X 1 ] + α 2 E [X 2 ] + · · · + α n E [X n ] = 0 41 (3)
.4 (A) Each K i has MGF φ K (s) = E es K i = es (1 − ens ) es + e2s + · · · + ens = n n(1 − es ) ems (1 − ens )m n m (1 − es )m (1)
Since the sequence of K i is independent. Theorem 6.10 says that W is a Gaussian random variable.2(es + 2e2s + 3e3s + 4e4s ) ds Evaluating the derivative at s = 0 yields E [K ] = d φ K (s) ds = 0.Quiz 6.8 says the MGF of J is φ J (s) = (φ K (s))m = (2)
(B) Since the set of α j X j are independent Gaussian random variables. The ﬁrst derivative of φ K (s) is d φ K (s) = 0.2(es + 8e2s + 27e3s + 64e4s )
s=0
s=0
= 0. we need only ﬁnd the expected value and variance.2(1 + 2 + 3 + 4) = 2
s=0
(2)
(3)
To ﬁnd higher-order moments.2)esk = 0. Thus to ﬁnd the PDF of W . Theorem 6.2 1 + es + e2s + e3s + e4s
(1)
We ﬁnd the moments by taking derivatives.8
(4) (5) (6) (7)
= 0.2(es + 4e2s + 9e3s + 16e4s )
s=0
s=0
=6 = 20 = 70.3 The MGF of K is
4
φ K (s) = E es K ==
k=0
(0.2(es + 16e2s + 81e3s + 256e4s )
s=0
s=0
Quiz 6. we continue to take derivatives: E K2 = E K3 E K4 d 2 φ K (s) ds 2 d 3 φ K (s) = ds 3 d 4 φ K (s) = ds 4 = 0.

1−s φ N (s) =
1 s 5e .6 to write Var[W ] = α 2 − α 2n+2 [1 + n(1 − α 2 )] (1 − α 2 )2 (6) (4) (5)
2 With E[W ] = 0 and σW = Var[W ].
42
.1. R has MGF φ R (s) = φ N (ln φ X (s)) = Substituting the expression for φ X (s) yields φ R (s) =
1 5 1 5 1 5 φ X (s) 1 − 4 φ X (s) 5
(2)
−s
. we can use Math Fact B.
(3)
(2) From Table 6. The corresponding PDF is f R (r ) = (1/5)e−r/5 r ≥ 0 0 otherwise (4)
This quiz is an example of the general result that a geometric sum of exponential random variables is an exponential random variable. we see that R has the MGF of an exponential (1/5) random variable.Since the α j X j are independent. the variance of the sum equals the sum of the variances: Var[W ] = α 2 Var[X 1 ] + α 4 Var[X 2 ] + · · · + α 2n Var[X n ] = α 2 + 2(α 2 )2 + 3(α 2 )3 + · · · + n(α 2 )n Deﬁning q = α 2 .1.12.5 (1) From Table 6. 1 − 4 es 5
(1)
From Theorem 6. we can write the PDF of W as
f W (w) =
1
2 2π σW
e−w
2 /2σ 2 W
(7)
Quiz 6. each X i has MGF φ X (s) and random variable N has MGF φ N (s) where φ X (s) = 1 .

the standard deviation of A is σ A = 12 (5) To use the central limit theorem. we use the central limit theorem and Table 3. (3) Using X i to denote the access time of block i.1 to look up (0. we can write A = X 1 + X 2 + · · · + X 12 Since the expectation of the sum equals the sum of the expectations. Var[A] = Var[X 1 ] + · · · + Var[X 12 ] = 12 Var[X ] = 144 Hence. (6) (7) (8) (9) (5) (4) (3)
(6) Once again.Quiz 6. E [A] = E [X 1 ] + · · · + E [X 12 ] = 12E [X ] = 72 msec (4) Since the X i are independent.25).9773 = 0.0227 (10) (11) (12)
43
.6 (1) The expected access time is E [X ] =
∞ −∞
x f X (x) d x =
0
12
x d x = 6 msec 12
(1)
(2) The second moment of the access time is E X2 =
∞ −∞
x 2 f X (x) d x =
0
12
x2 d x = 48 12
(2)
The variance of the access time is Var[X ] = E[X 2 ] − (E[X ])2 = 48 − 36 = 12.1 to estimate P [A < 48] = P 48 − E [A] A − E [A] < σA σA 48 − 72 ≈ 12 = 1 − (2) = 1 − 0.5987 = 0.4013 Note that we used Table 3. we write P [A > 75] = 1 − P [A ≤ 75] 75 − E [A] A − E [A] ≤ =1− P σA σA 75 − 72 ≈1− 12 = 1 − 0.

we ﬁnd that W has expected value and variance E [W ] = 3/λ = 6 Var[W ] = 3/λ2 = 12 (2) (1) By the Central Limit Theorem.5 − 36 30 − 0. X 2 .1 yields P [30 ≤ K 48 ≤ 42] ≈ Recalling that (−x) = 1 − 42 − 36 − 3 (x). we can use the De Moivre-Laplace approximation to estimate P [30 ≤ K 48 ≤ 42] ≈ 42 + 0. (3) Using the ordinary central limit theorem and Table 3. (1)
In Theorem 6. P [W > 20] = P √ W −6 20 − 6 > √ ≈ Q(7/ 3) = 2.66 × 10−5 √ 12 12 (3)
44
. (2) The variance of K 48 is Var[K 48 ] = 48P [V ] (1 − P [V ]) = 48(3/4)(1/4) = 9 Thus K 48 has standard deviation σ K 48 = 3.9687 (4) (5)
Quiz 6.9545
(4) Since K 48 is a discrete random variable.11.8 The train interarrival times X 1 . From Appendix A. (1) The expected number of voice calls out of 48 calls is E[K 48 ] = 48P[V ] = 36.Quiz 6.7 Random variable K n has a binomial distribution for n trials and success probability P[V ] = 3/4. The arrival time of the third train is W = X 1 + X 2 + X 3. we found that the sum of three iid exponential (λ) random variables is an Erlang (n = 3. we have (3) 30 − 36 3 = (2) − (−2) (2) (1)
P [30 ≤ K 48 ≤ 42] ≈ 2 (2) − 1 = 0. X 3 are iid exponential (λ) random variables.16666) − 1 = 0.5 − 36 − 3 3 = 2 (2. λ) random variable.

A graph of the PMF PW (w) appears in Figure 2 With some thought. P [W > 20] = 1 − FW (20) = e−10 1 + 10 102 + 1! 2! = 61e−10 = 0.5.sy).19:
%unifbinom100.sy=0:100. By contrast.0. Applying s = 7/20 into the Chernoff bound yields P [W > 20] ≤ e−20s (1 − 2s)3 = (10/3)3 e−7 = 0. Quiz 6.SY]=ndgrid(sx. we note that the MGF of W is φW (s) = The Chernoff bound states that P [W > 20] ≤ min e−20s φ X (s) = min
s≥0 s≥0
λ λ−s
3
=
1 (1 − 2s)3 e−20s (1 − 2s)3
(4)
(5)
To minimize h(s) = e−20s /(1 − 2s)3 . pmfplot(sw.’\itP_W(w)’).py).’\itw’. 45
. we set the derivative of h(s) to zero: −20(1 − 2s)3 e−20s + 6e−20s (1 − 2s)2 d h(s) = =0 ds (1 − 2s)6 (6)
This implies 20(1 − 2s) = 6 or s = 7/20. the Central Limit Theorem approximation grossly underestimates the true probability.100.pw. it should be apparent that the finitepmf function is implementing the convolution of the two PMFs. py=duniformpmf(0. PW=PX. SW=SX+SY. sw=unique(SW).0028 (9) (10)
Although the Chernoff bound is relatively weak in that it overestimates the probability by roughly a factor of 12.sx).sw). the CDF of the Erlang (λ. it is a valid bound.(2) To use the Chernoff bound.PW.*PY.0338
s=7/20
(7)
(3) Theorem 3.m sx=0:100. [SX. [PX.PY]=ndgrid(px. 3) random variable W satisﬁes 2 (λw)k e−λw FW (w) = 1 − (8) k!
k=0
Equivalently.sy). pw=finitepmf(SW.9 One solution to this problem is to follow the approach of Example 6. px=binomialpmf(100. for λ = 1/2 and w = 20.11 says that for any w > 0.

008 PW(w) 0.9.
46
.006 0. 0. the PMF PW (w) of the independent sum of a binomial (100.002 0 0 20 40 60 80 100 w 120 140 160 180 200
Figure 2: From Quiz 6.01 0.5) random variable and a discrete uniform (0.004 0. 100) random variable.0.

12 Thus E[W ] = 3E[X i ] = 45. (1) By the Markov inequality. By Theorem 7. we need n = 100 samples. Thus.1. Mn (X ) has variance Var[Mn (X )] = 1/n.6. (30 − 0)2 Var [X i ] = = 75. Hence. µW = E X
2
Var[W ] 100
(1)
=
1 −1 1 −1
x 2 f X (x) d x = 1/3 x 4 f X (x) d x = 1/5
(2) (3)
E W2 = E X4 =
Therefore Var[W ] = E[W 2 ] − µ2 = 1/5 − (1/3)2 = 4/45 and the mean square error is W 4/4500 = 0.000889. and Var[W ] = 3 Var[X i ] = 225. Observe that V100 (X ) = M100 (W ).Quiz Solutions – Chapter 7
Quiz 7. Quiz 7.2 The arrival time of the third elevator is W = X 1 + X 2 + X 3 . By Theorem 7. the mean square error is E (M100 (W ) − µW )2 = Observe that µ X = 0 so that W = X 2 . Since each X i is uniform (0. 30). P [W > 75] ≤ (2) By the Chebyshev inequality.
47
.1 An exponential random variable with expected value 1 also has variance 1. (1) E [X i ] = 15. P [W > 75] = P [W − E [W ] > 30] ≤ P [|W − E [W ]| > 30] ≤ 225 Var [W ] 1 = = 2 900 4 30 (3) (4) E [W ] 45 3 = = 75 75 5 (2)
Quiz 7.3 Deﬁne the random variable W = (X − µ X )2 .

m generates graphs the number of traces within one standard error as a function of the time. we must have √ c n ≥ 0. In this case. OK(k) counts the fraction of sample paths that have sample mean within one standard error of p. p(1 − p)
(2)
We must ensure for every value of p that 1 − α ≥ 0. Since p(1 − p) ≤ 1/4 for all p. we can use a Gaussian approximation for Mn (X ).5 Following the approach of bernoullitraces.e.3355 ≤ p ≤ 0. Equivalently. implying (c n/( p(1− p))) ≥ 0.99 conﬁdence is high. we generate m = 1000 sample paths. √ This implies c n√ 2. The interval is wide because the 0.1.9 or α ≤ 0. SinceE[X ] = p and Var[X ] = p(1 − p). 4 n n The 0. The program bernoullisample.4645. Since p(1 − p) ≤ 1/4 for all p. we have α ≤ 0. i.58)/ n.58 p(1 − p). we require that ≥ c ≥ (0. Quiz 7.99 conﬁdence interval estimate is 0. the number of trials in each trace.65 p(1 − p).4.645 0. the 0. Since (x) is an increasing function of x.65 0.13 which says that the interval estimate Mn (X ) − c ≤ p ≤ Mn (X ) + c (1) has conﬁdence coefﬁcient 1 − α where α =2−2 √ c n . 48 (7) (6)
. each sample path having n = 100 Bernoulli traces. at time k.995. n n Note that if M100 (X ) = 0.9 conﬁdence interval estimate of p is 0. we must satisfy c n ≥ 1.99 conﬁdence interval estimate is 0.645 Mn (X ) − √ ≤ p ≤ Mn (X ) + √ .m.99 conﬁdence interval.25)(2.41 Mn (X ) − √ ≤ p ≤ Mn (X ) + √ .41 0. then the 0. we require that 1.95 (3) p(1 − p) √ for every value of p.41 c≥ √ = √ . we apply Theorem 7.4 Assuming the number n of samples is large.01.Quiz 7. n n (5) (4)
√ For the 0.

5000./nn. nn=(1:n)’*ones(1. though perhaps unexpected. the fraction of traces within one standard error approaches 2 (1) − 1 ≈ 0.n.5):
1 0.function OK=bernoullisample(n.5 0.
The following graph was generated by bernoullisample(100.m).
49
.p). x=reshape(bernoullirv(p. OK=sum(abs(MN-p)<stderrmat. The unusual sawtooth pattern. as m gets large. plot(1:n. is examined in Problem 7.4 0 10 20 30 40 50 60 70 80 90 100
As we would expect.2. stderrmat=stderr*ones(1.m.68.5.9 0.8 0.OK.2)/m.’-s’).m).6 0.0.7 0. stderr=sqrt(p*(1-p)).m*n)./sqrt((1:n)’).m). MN=cumsum(x).

otherwise k = 0. FX (x) = FX i (x)
15
(2)
= 1 − e−x
15
(3)
To design a signiﬁcance test. A reasonable choice is to reject the hypothesis if X is too small. This rule simpliﬁes to 106 − 104 k ∈ A0 if k ≤ k = = 214. This implies that for x ≥ 0.7.33. X 15 obeys FX (x) = P [X ≤ x] = P [X 1 ≤ x. 1.01. otherwise
(1) (2)
0
Since the two hypotheses are equally likely. .01)1/15 = 1. . the conditional PMFs of K are PK |H0 (k) = PK |H1 (k) =
104k e−10 k!
4
(4)
(5)
0
106k e−10 k!
6
k = 0. . . 975. . then we reject the hypothesis. we obtain α = P [X ≤ r ] = (1 − e−r )15 = 0.2 From the problem statement.01 It is straightforward to show that r = − ln 1 − (0. if we observe X < 1. the CDF of the maximum of X 1 .6. X 15 ≤ x] = [P [X i ≤ x]]15 . the ML hypothesis rule is k ∈ A0 if PK |H0 (k) ≥ PK |H1 (k) . we must choose a rejection region for X . 976 photons.33 Hence. 50
.
(3)
k ∈ A1 otherwise. . For a signiﬁcance level of α = 0.1 From the problem statement. then we accept hypothesis H1 . ln 100
∗
k ∈ A1 otherwise.
(4)
Thus if we observe at least 214. let R = {X ≤ r }. each X i has PDF and CDF f X i (x) = e−x x ≥ 0 0 otherwise FX i (x) = 0 x <0 1 − e−x x ≥ 0 (1)
Hence. That is. From Theorem 8. . · · · .Quiz Solutions – Chapter 8
Quiz 8. . the MAP and ML tests are the same. 1. . X 2 ≤ x. . Quiz 8.

FM5(:.1) %add d(v+N)ˆ2 distortion %receive 1 if x>T.1).’-k’.T(:)). FM5(:.1). X 2 ) ∈ A j for some j = i.T). .. P10=sum((XX+d*(XX.3 For the QPSK system.T).m. xlabel(’P_{FA}’).m. the program sqdistrocplot.m. Since N1 and N2 are iid Gaussian (0.1). Equivalently.m calls sqdistroc three times to generate a plot that compares the receiver performance for the three requested values of d.2’.m. function FM=sqdistrocplot(v.’:k’). the probability
2
PERR = 1 − P [C] = 1 −
E 2σ 2
(5)
Quiz 8.ˆ2)< TT)... .T).T).1)/m.0. For a QPSK system.2. loglog(FM1(:.1). The modiﬁed program. the existing program sqdistor already calculates this miss probability PMISS = P01 and the false alarm probability PFA = P10 . FM2(:. FM=[FM1 FM2 FM5]. sqdistroc.4 To generate the ROC. FM=[P10(:) P01(:)]. FM1=sqdistroc(v. x= -v+randn(m. [XX.m. otherwise 0 %FM = [P(FA) P(MISS)] x=(v+randn(m.2). it is easier to calculate the probability of a correct decision. the conditional probability of a correct decision is √ √ P [C|H0 ] = P [X 1 > 0.3) ylabel(’P_{MISS}’). a symbol error occurs when si is transmitted but (X 1 .0. FM2=sqdistroc(v.2). FM5=sqdistroc(v.FM1(:.m is essentially the same as sqdistor except the output is a matrix FM whose columns are the false alarm and miss probabilities.1’.1)).ˆ2)>TT).TT]=ndgrid(x. [XX.1)/m.2). This implies the probability of a correct decision is P[C] = P[C|H0 ]..1.. Given H0 .. we have √ √ P [C] = P [C|H0 ] = P E/2 + N1 > 0 P E/2 + N2 > 0 (2) √ 2 (3) = P N1 > − E/2 √ 2 − E/2 (4) = 1− σ Since (−x) = 1 − of error is (x).T) %square law distortion recvr %P(error) for m bits tested %transmit v volts or -v volts. legend(’\it d=0.3’. we have P[C] =
2(
E/2σ 2 ).3.
51
. X 2 > 0|H0 ] = P E/2 + N1 > 0.0. σ ) random variables. Next. Here is the modiﬁed code:
function FM=sqdistroc(v.TT]=ndgrid(x.T(:)). ’\it d=0.FM2(:. %add N volts.’--k’..’\it d=0. P[C|H0 ] = P[C|Hi ] for all i.d. N is Gauss(0.Quiz 8. E/2 + N2 > 0 (1) Because of the symmetry of the signals. P01=sum((XX+d*(XX.

Figure 3: The receiver operating curve for the communications system of Quiz 8. the commands T=-3:0.T).1:3.
52
. sqdistrocplot(3.3
−5
10
10
−4
10
−3
10 PFA
−2
10
−1
10
0
T=-3:0.
10
0
10
−1
10 PMISS 10 10
−2
−3
−4
10
−5
d=0.100000.1 d=0.2 d=0. sqdistrocplot(3.To see the effect of d.4 with squared distortion.100000. generated the plot shown in Figure 3.1:3.T).

Y (x. f X (x) =
x 1
2(y + x) dy = y 2 + 2x y
y=1 y=x
= 1 + 2x − 3x 2
(4) (5)
For 0 ≤ x ≤ 1. the conditional PDF of Y given X is f Y |X (y|x) =
2(y+x) 1+2x−3x 2
0
x ≤y≤1 otherwise
(6)
(4) The MMSE estimate of Y given X = x is y M (x) = E [Y |X = x] = ˆ
x 1
2y 2 + 2x y dy 1 + 2x − 3x 2
y=1 y=x
(7) (8) (9)
2y 3 /3 + x y 2 = 1 + 2x − 3x 2 = 2 + 3x − 5x 3 3 + 6x − 9x 2
53
. we calculate the marginal PDF for 0 ≤ y ≤ 1: f Y (y) =
0 y
2(y + x) d x = 2x y + x 2
x=y x=0
= 3y 2
(1)
This implies the conditional PDF of X given Y is f X |Y (x|y) = f X.1 (1) First. For 0 ≤ x ≤ 1.Quiz Solutions – Chapter 9
Quiz 9. (3) To obtain the conditional PDF f Y |X (y|x). y) = f Y (y)
2 3y
+
2x 3y 2
0
0≤x ≤y otherwise
(2)
(2) The minimum mean square error estimate of X given Y = y is x M (y) = E [X |Y = y] = ˆ
0 y
2x 2 2x + 2 3y 3y
d x = 5y/9
(3)
ˆ Thus the MMSE estimator of X given Y is X M (Y ) = 5Y /9. we need the marginal PDF f X (x).

4. E [R] = E [T ] + E [X ] = 0 (2) Since T and X are independent.R = σT /σ R .8. the correlation coefﬁcient of T and R is ρT.R = √ √ σT Cov [T. the conditional PDF of X = Y −40−40 log10 r is Gaussian with expected value −40 − 40 log10 r and variance 64. (4) From Deﬁnition 4. E[T X ] = E[T ]E[X ] = 0 and E[T 2 ] = Var[T ].2 (1) Since the expectation of the sum equals the sum of the expectations. The conditional PDF of X given R is 1 2 f X |R (x|r ) = √ e−(x+40+40 log10 r ) /128 128π 54 (1)
. R] = = 3/2 σR Var[R] Var[T ] (4)
(5) From Theorem 9.3 When R = r . the mean square error of the linear estimate is
2 e∗ = Var[T ](1 − ρT. R] = Var[T ] = 9. ˆ TL (R) = Hence a ∗ = 3/4 and b∗ = 0. R] = E [T R] = E [T (T + X )] = E T 2 + E [T X ] (3) (2) (1)
Since T and X are independent and have zero expected value.4.R (R − E [R]) + E [T ] σR Since E[R] = E[T ] = 0 and ρT. Cov [T. (6) By Theorem 9.R ) = 9(1 − 3/4) = 9/4 L 2 σT
(5)
σR
R= 2
2 σT 2 2 σT + σ X
R=
3 R 4
(6)
(7)
Quiz 9.Quiz 9. the optimum linear estimate of T given R is σT ˆ TL (R) = ρT. Thus Cov[T. the variance of the sum R = T + X is Var[R] = Var[T ] + Var[X ] = 9 + 3 = 12 (3) Since T and R have expected values E[R] = E[T ] = 0.

However. This minimum occurs when the exponent is zero.From the conditional PDF f X |R (x|r ). r ) = f X |R (x|r ) f R (r ) = 106 32π 1 √ r e−(x+40+40 log10 r )
2 /128
(5)
From Theorem 9. 55
.R (x.3 dB.2 to write the ML estimate of R given X = x as rML (x) = arg max f X |R (x|r ) ˆ
r ≥0
(2)
We observe that f X |R (x|r ) is maximized when the exponent (x + 40 + 40 log10 r )2 is minimized. When x ≤ −156. r ).R (x. if x = −120dB.3 (0.6. This corresponds to a distance estimate of rML (−120) = 100 m. the complete description of the MAP estimate is rMAP (x) = ˆ 1000 x < −156. That is. we can use Deﬁnition 9. This reﬂects the fact that large values of R are a priori more probable than small values. for very low signal strengths. R ≤ 1000 m.1236)10−x/40
(8)
This is the MAP estimate of R given X = x as long as r ≤ 1000 m. the MAP estimate is 23. the MAP estimate takes into account that the distance can never exceed 1000 m. the MAP estimate of R given X = x is the value of r that maximizes f X. ˆ For the MAP estimate. note that a typical ﬁgure for the signal strength might be x = −120 dB. then rMAP (−120) = 123. yielding log10 r = −1 − x/40 or rML (x) = (0. which is not possible in our probability model.R (x.6% larger than the ML estimate.1236)10 (9)
For example.6 m. Hence. Setting the derivative of f X. the above estimate will exceed 1000 m. When the measured signal ˆ strength is not too low.3 −x/40 x ≥ −156. r ) ˆ
0≤r ≤1000
Note that we have included the constraint r ≤ 1000 in the maximization to highlight the fact that under our probability model. (6) rMAP (x) = arg max f X.R (x.1)10−x/40 m ˆ (3) (4)
If the result doesn’t look correct. we observe that the joint PDF of X and R is f X. r ) with respect to r to zero yields e−(x+40+40 log10 r ) Solving for r yields r = 10
1 25 log10 e −1 2 /128
1−
80 log10 e (x + 40 + 40 log10 r ) = 0 128
(7)
10−x/40 = (0.

to compute the expected square error. it follows that E[Y] = 0. To apply Theorem 9.1 (6)
In terms of Theorem 9. This implies RY = E XX + E WW = RX + RW = In addition. we need to ﬁnd RY and RYX 2 . we need to ﬁnd RYX 2 = E [YX 2 ] = E [Y1 X 2 ] E [(X 1 + W1 )X 2 ] = .7.Y2 ) = 1 − L
Cov [X 2 . it follows that b∗ = 0. 0 0. E [Y2 X 2 ] E [(X 2 + W2 )X 2 ] 56 (10) 1.1 11
(5)
(2) Since Y = X + W and E[X] = E[W] = 0. n = 2 and we wish to estimate X 2 given the observation vector Y = Y1 Y2 .1 (9)
. RY = E YY = E (X + W)(X + W ) = E XX + XW + WX + WW . the LMSE estimate of X 2 given Y2 is X 2 (Y2 ) = a ∗ Y2 + b∗ where a∗ = Cov [X 2 . (1)
Because E[X] = E[Y] = 0. Similarly.7.1 0 .
2 Cov [X 2 . E[XW ] = E[X]E[W ] = 0.7.4.Quiz 9.1
(2) (3)
It follows that a ∗ = 1/1. Thus we can apply Theorem 9.4 ˆ (1) From Theorem 9. we calculate the correlation coefﬁcient ρ X 2 .0909 1.1. Var[Y2 ] b ∗ = µ X 2 − a ∗ µ Y2 . Note that X and W have correlation matrices RX = 1 −0.1 −0.9 . −0.9 1 RW = 0. Because µ X 2 = µY2 = 0. Y2 ] = E [X 2 Y2 ] = E [X 2 (X 2 + W2 )] = E X 2 = 1 2 2 Var[Y2 ] = Var[X 2 ] + Var[W2 ] = E X 2 + E W2 = 1.9 1. Finally. E[WX ] = 0. −0.1
(4)
1 1 = = 0.9 . Y2 ] . Y2 ] 1 =√ σ X 2 σY2 1. (7) (8)
Because X and W are independent.Y2 = The expected square error is
2 e∗ = Var[X 2 ](1 − ρ X 2 .

ˆ a = R−1 RYX = 11 + RW Y and the optimal linear estimator is ˆ X L (Y) = 1 11 + RW The mean square error is ˆ e∗ = Var[X ] − a RYX = 1 − 1 11 + RW L
−1 −1 −1
(1)
(2) (3) (4)
1
(5)
Y
(6)
1
(7)
Now we note that RW has i.0725. By the same reasoning. This implies RYX = E [YX ] = E [(1X + W)X ] = 1E X 2 = 1. Thus.7. Thus. The question we must address is what value c minimizes e∗ . (14) (13)
Quiz 9. the correlation matrix of Y is RY = E YY = E (1X + W)(1 X + W ) = 11 E X 2 + 1E X W + E [WX ] 1 + E WW = 11 + RW Note that 11 is a 20 × 20 matrix with every entry equal to 1. Y E[WX ] = 0 and E[X W ] = 0 .X 2 = 0. j) = c|i− j|−1 . the optimum linear estimator of X 2 given Y1 and Y2 is ˆ ˆ X L = a Y = −0.7. This problem is atypical in that one does not usually get L 57
. by ˆ ˆ ˆ Theorem 9. X L (Y) = a Y where a = R−1 RYX . The mean square error is ˆ Var [X 2 ] − a RYX 2 = Var [X ] − a1rY1 . Thus E[X 1 X 2 ] −0.725 (12)
Therefore. ˆ a = R−1 RYX 2 = Y −0.725Y2 . (11) 2 1 E X2 By Theorem 9.225Y1 + 0. Y also has zero expected value.X 2 − a2rY2 .5 Since X and W have zero expected value.9 RYX 2 = = . E[W1 X 2 ] = E[W1 ]E[X 2 ] = 0 and E[W2 X 2 ] = 0. jth entry RW (i. Since X and W are independent.225 0.Since X and W are independent vectors.

Note in mquiz9 that v1 corresponds to the vector 1 of all ones.4500
1 0. msec=zeros(size(c)).01:0. However. Thus. RW=toeplitz(c. we observe that Var[Wi ] = RW (i. function cmin=mquiz9minc(c).af]=mquiz9(c(k)).6 0.ˆ((0:19)-1)). xlabel(’c’).to choose the correlation structure of the noise. both small values and large values of c result in large MSE. we write a M ATLAB function mquiz9(c) to calculate the MSE for a given c and second function that ﬁnds plots the MSE for a range of values of c. [msec(k).af]=mquiz9(c).optk]=min(msec). the noises Wi have high variance and we would expect our estimator to be poor. cmin=c(optk). To ﬁnd the optimal value of c. We note that the answer is not obviously apparent from Equation (7).01:0. In particular.ylabel(’e_Lˆ*’).msec). consider the extreme case in which every Wi and W j have correlation coefﬁcient ρi j = 1. This would suggest that large values of c will also result in poor MSE.
58
. In this case.1). we will see that the answer is somewhat instructive. af=(inv(RY))*v1. when c is small.4 0. our 20 measurements will be all the same and one measurement is as good as 20 measurements. if c is large Wi and W j are highly correlated and the separate measurements of X are very dependent.2 0 0.5 c 1
As we see in the graph. mse=1-((v1’)*af). i) = 1/c.
function [mse. for k=1:length(c). >> mquiz9minc(c) ans = 0. If this argument is not clear. The following commands ﬁnds the minimum c and also produces the following graph:
>> c=0.99. v1=ones(20. end plot(c. On the other hand.8 e* L 0. RY=(v1*(v1’)) +RW. [msemin.

the number of new calls that arrive during the experiment • X 1 . . s) is • m(0. . continuous valued process. we round the temperature to the nearest degree. discrete valued process.01) dr = 0.2
(2)
59
. . the number of calls that hang up during the experiment • D1 .2 (1) We obtain a continuous time. . . then we obtain a continuous time. the call completion times of the H calls that hang up Quiz 10. discrete valued process.3 (1) Each resistor has resistance R in ohms with uniform PDF f R (r ) = 0. (3) If we sample the process in part (a) every T seconds. then we obtain a discrete time. the number of ongoing calls at the start of the experiment • N .01 950 ≤ r ≤ 1050 0 otherwise (1)
The probability that a test produces a 1% resistor is p = P [990 ≤ R ≤ 1010] =
1010 990
(0. Quiz 10. continuous valued process when we record the temperature as a continuous waveform over time.Quiz Solutions – Chapter 10
Quiz 10. A correct answer speciﬁes enough random variables to specify the sample path exactly. D H .1 There are many correct answers to this question. One choice for an alternate set of random variables that would specify m(t. . (4) Rounding the samples in part (c) to the nearest integer degree yields a discrete time. . s). . X N . the interarrival times of the N new arrivals • H . (2) If at every moment in time.

That is. This problem is easy if we view each resistor test as an independent trial.(2) In t seconds. A success occurs on a trial with probability p if we ﬁnd a 1% resistor..5.2) = 0. .08192. .8)4 (0. .1. Thus E [T2 |T1 = 10] = E [T1 |T1 = 10] + E T |T1 = 10 = 10 + E T = 10 + 5 = 15 (5) (6)
Quiz 10.. . the probability the ﬁrst 1% resistor is found in exactly ﬁve seconds is PT1 (5) = (0. . T2 = T1 + T where T is independent and identically distributed to T1 .11. (5) Note that once we ﬁnd the ﬁrst 1% resistor.2. a geometric random variable with success probability p has expected value 1/ p. In this problem. E[T1 ] = 1/ p = 5. t − 1 followed by a success on trial t. Consequently. each X i has PDF 1 2 f X (i) (x) = √ e−x /2 2π By Theorem 10. t 0 otherwise
t n
(3)
(3) First we will ﬁnd the PMF of T1 . .4 Since each X i is a N (0. Hence. the number of additional trials needed to ﬁnd the second 1% resistor once again has a geometric PMF with expected value 1/ p since each independent trial is a success with probability p. 1. independent of any other resistor. 9 otherwise (4)
Since p = 0.. . Each resistor is a 1% resistor with probability p. . The ﬁrst 1% resistor is found at time T1 = t if we observe failures on trials 1. (4) From Theorem 2. just as in Example 2. the joint PDF of X = X 1 · · · X n is
k
(1)
f X (x) = f X (1). xn ) =
i=1
f X (xi ) =
1 2 2 e−(x1 +···+xn )/2 n/2 (2π )
(2)
60
. . . . the number of 1% resistors found has the binomial PMF PN (t) (n) = p n (1 − p)t−n n = 0. exactly t resistors are tested. . . .X (n) (x1 . 1) random variable. T1 has the geometric PMF PT1 (t) = (1 − p)t−1 p t = 1.. 2. .

Since we count only evennumbered arrival for N (t).6 To answer whether N (t) is a Poisson process.11. Let X 1 . the joint PMF of M1 and M2 is ⎧ α m 1 +m 2 e−2α m 1 = 0. we note that for t > s. . . . . we can conclude that the interarrival times of N (t) are not exponential random variables. . 000. 61
.
(2)
Quiz 10. This implies < √ [W (t) − W (s)]/ α is independent of W (s )/ α for all s ≥ s .Quiz 10.7 First. denote the interarrival times of the N (t) process. Since one hour equals 3600 sec and the Poisson process has a rate of 10 packets/sec. . the expected number of packets in each hour is E[Mi ] = α = 36. W (t) − W (s) is independent of W (s ). Thus N (t) is not a Poisson process. Since Yi (t). 1. 1. otherwise
(1)
Since M1 and M2 are independent. ⎪ m 1 !m 2 ! ⎪ ⎨ m 2 = 0. the ith interarrival time of the N (t) process. 2. This implies M1 and M2 are independent Poisson random variables each with PMF PMi (m) =
α m e−α m!
0
m = 0. . . Y1 is an Erlang (n = 2. see Theorem 6. λ) random variable. . . That is. Since s ≥ s . Since X 1 and X 2 are independent exponential (λ) random variables. has the same PDF as Y1 (t). X (t) − X (s) is independent of X (s ) for all s ≥ s . .5 The ﬁrst and second hours are nonoverlapping intervals. X (t) − X (s) = W (t) − W (s) √ α (1)
Since W (t) − W (s) is a Gaussian random variable. X 2 . .M2 (m 1 .13 states that W (t) − W (s) is Gaussian with expected value E [X (t) − X (s)] = and variance E (W (t) − W (s))2 = E (W (t) − W (s))2 α(t − s) = α α (3) E [W (t) − W (s)] =0 √ α (2)
Consider s ≤ s √ t. Theorem 3. m 2 ) = PM1 (m 1 ) PM2 (m 2 ) = ⎪ ⎪ ⎩ 0 otherwise. Thus X (t) is a Brownian motion process with variance Var[X (t)] = t. the time until the ﬁrst arrival of the N (t) is Y1 = X 1 + X 2 . . we look at the interarrival times. Quiz 10. PM1 . 1. .

. X 2 . . . . (2) R2 (τ ) = e−τ also is valid... .14. τ ) = E[Y (t)Y (t + τ )]. . .. is a stationary random sequence if for all sets of time instants n 1 .10 We must check whether each function R(τ ) meets the conditions of Theorem 10..Quiz 10..X nm (x1 .X nm +k (x1 . f X n1 . X 1 .. . n m + k. . . xm ) = f X (x1 ) f X (x2 ) · · · f X (xm ) Similarly.
2
(3) R3 (τ ) = e−τ cos τ is not valid because R3 (−2π ) = e2π cos 2π = e2π > 1 = R3 (0) (4) R4 (τ ) = e−τ sin τ also cannot be an autocorrelation function because
2
(2)
R4 (π/2) = e−π/2 sin π/2 = e−π/2 > 0 = R4 (0)
(3)
62
. f X n1 +k .. f X n1 . n m and time offset k.9 From Deﬁnition 10.. .. we observe that since X (t) and N (t) are independent and since N (t) has zero expected value. for time instants n 1 + k. . .. xm ) = f X n1 +k . .X nm +k (x1 . Since RY (t.. (1)
To ﬁnd the autocorrelation. . . τ ) = E [(X (t) + N (t)) (X (t + τ ) + N (t + τ ))] = E [X (t)X (t + τ )] + E [X (t)N (t + τ )] + E [X (t + τ )N (t)] + E [N (t)N (t + τ )] = R X (t. E[X (t)N (t )] = E[X (t)]E[N (t )] = 0. . . we have RY (t.X nm (x1 . . Quiz 10. .. .8 First we ﬁnd the expected value µY (t) = µ X (t) + µ N (t) = µ X (t).. τ ) + R N (t.. . (2) (3) (4)
Quiz 10. τ ). .. .. . xm ) = f X (x1 ) f X (x2 ) · · · f X (xm ) We can conclude that the iid random sequence is stationary. . .12: R(τ ) ≥ 0 R(τ ) = R(−τ ) |R(τ )| ≤ R(0) (1) (3) (2) (1)
(1) R1 (τ ) = e−|τ | meets all three conditions and thus is valid. xm ) Since the random sequence is iid.

τ ) = E [Y (t)Y (t + τ )] = E [X (−t)X (−t − τ )] = R X (−t − (−t − τ )) = R X (τ ) (1) (2) (3)
Since E[Y (t)] = E[X (−t)] = µ X . Quiz 10. we can check whether they are jointly wide sense stationary by seeing if R X Y (t. τ ) depends on both t and τ .11 (1) The autocorrelation of Y (t) is RY (t. R X Y (t. E [X (t)] = E [X (t + 1)] = 0 E [X (t)X (t + 1)] = 1/2 Var[X (t)] = Var[X (t + 1)] = 1 The Gaussian random vector X = X (t) X (t + 1) sponding inverse CX = Since 1 1/2 1/2 1 C−1 = X (1) (2) (3)
has covariance matrix and corre-
4 1 −1/2 1 3 −1/2
(4)
4 4 2 1 −1/2 x0 2 x − x0 x+ x1 = 1 x1 3 −1/2 3 0 the joint PDF of X (t) and X (t + 1) is the Gaussian vector PDF x C−1 x = x0 x1 X f X (t). τ ) = E [X (t)Y (t + τ )] = E [X (t)X (−t − τ )] = R X (t − (−t − τ )) = R X (2t + τ ) (4) (5) (6)
Since R X Y (t. x1 ) = 1 (2π )n/2 [det (CX )]1/2 1 3π 2 e− 3
2 2 2 x0 −x0 x1 +x1
(5)
1 exp − x C−1 x X 2
(6) (7)
=√
63
. as t gets larger.Quiz 10. Y (t) = X (−t) and X (t) become less and less correlated. (2) Since X (t) and Y (t) are both wide sense stationary processes. we see that by viewing a process backwards in time.12 From the problem statement. In fact. In this case.X (t+1) (x0 . we conclude that X (t) and Y (t) are not jointly wide sense stationary. suppose R X (τ ) = e−|τ | so that samples of X (t) far apart in time have almost no correlation. τ ) is just a function of τ . we see the same second order statistics. To see why this is. In this case. we can conclude that Y (t) is a wide sense stationary process.

The logic of such a simulation is 1. The program simply executes the event at the head of the schedule.13. and schedule a departure to occur at time t + Sn . the number of ongoing calls. 3. • If the head of schedule event is a departure. 2. increase the system state n by 1. an exponential (λ) random variable. Examine the head-of-schedule event. when an arrival occurs at time t. we cannot generate these vectors all at once. In particular. Otherwise. admit the arrival. satisﬁes M(t) < c = 120. we must block the call. Start at time t = 0 with an empty system. we need to know that M(t). With the introduction of call blocking. Quiz 10. when M(t) = c. – If M(t) = c.28 admits a deceptively simple solution in terms of the vector of arrivals A and the vector of departures D. After the head-of-schedule event is completed and any new events (departures in this system) are scheduled. – If M(t) < c. • When the head-of-schedule event is the kth arrival is at time t. check the state M(t). Delete the head-of-schedule event and go to step 2. do not schedule a departure event. at discrete time instances.120 100 80 M(t) 60 40 20 0 0 10 20 30 40 50 t 60 70 80 90 100
Figure 4: Sample path of 100 minutes of the blocking switch of Quiz 10.13 The simple structure of the switch simulation of Example 10. block the arrival. we know the system state cannot change until the next scheduled event. The blocking switch is an example of a discrete event system. A simulation of the system moves from one time instant to the next by maintaining a chronological schedule of future events (arrivals and departures) to be executed. namely arrivals and departures. where Sk is an exponential (λ) random variable. 64
. The system evolves via a sequence of discrete events. Schedule the ﬁrst arrival to occur at S1 . Call blocking can be implemented by setting the service time of the call to zero so that the call departs as soon as it arrives. reduce the system state n by 1.

However. Thus this would account for only part of the disparity.000 minute full simulation produced a=49658 admitted calls and b=239 blocked calls. a simple (but not elegant) way to do this is to have maintain two vectors: time is a list of timestamps of scheduled events and event is a the list of event types.
65
. In our simulation.
generated a simulation lasting 5.” From the Erlang-B formula. Note that in Chapter 12. When the program is passed a vector t. We can estimate the probability a call is blocked as b ˆ = 0.0057 is that a simulation that includes only 239 blocks is not all that likely to give a very accurate result for the blocking probability. The complete program is shown in Figure 5.0048 and 0.b]=simblockswitch(10.000 minutes.1:5000. a kind of Markov chain.1. Thus for all times t(i) between the current head-of-schedule event and the next.0. In this case.m). or event(i)=-1 if the ith scheduled event is a departure. In most programming languages. The following instructions
t=0:0.0048. we will learn that the blocking switch is an example of an M/M/c/c queue. the output [m a b] is such that m(i) is the number of ongoing calls at time t(i) while a and b are the number of admits and blocks. we can calculate that the exact blocking probability is Pb = 0. One reason our simulation underestimates the blocking probability is that in a 5. for very complicated systems. Chapter 12 develops techniques for analyzing and simulating systems described by Markov chains that are much simpler than the discrete event simulation technique shown here. it is common to implement the event schedule as a linked list where each item in the list has a data structure indicating an event timestamp and the type of the event. (1) Pb = a+b In Chapter 12.t). plot(t. A sample path of the ﬁrst 100 minutes of that simulation is shown in Figure 4. the discrete event simulation is widely-used and often very efﬁcient simulation method. we use the vector t as the set of time instances at which we inspect the system state.a. event(i)=1 if the ith scheduled event is an arrival. In M ATLAB. we will learn that the exact blocking probability is given by Equation (12. roughly the ﬁrst 100 minutes are needed to load up the switch since the switch is idle when the simulation starts at time t = 0. [m. Nevertheless. we set m(i) to the current switch state.000 minute simulation.120.0057. The rest of the gap between 0. The 5.93). a result known as the “Erlang-B formula. this says that roughly the ﬁrst two percent of the simulation time was unusual.Thus we know that M(t) will stay the same until then.

n=n+1..blocks)).blocks]=simblockswitch(lam. %total # admits M=zeros(size(t)). eventnow=event(1). event=[ 1 ].13.admits. depart=timenow+exponentialrv(mu. immed departure disp(sprintf(’Time %10.mu.function [M. event=[event(b4depart) -1 event(˜b4depart)]. time=[time(b4depart) depart time(˜b4depart)].c. time(1)= [ ].1).1). event(1)=[ ]. timenow=time(1). %total # blocks admits=0. event=[event(b4arrival) 1 event(˜b4arrival)]. % # in system time=[ exponentialrv(lam.. if n<c %call admitted admits=admits+1.t).3d Admits %10d Blocks %10d’. tmax=max(t). timenow. else blocks=blocks+1. % next arrival b4arrival=time<arrival. end end
Figure 5: Discrete event simulation of the blocking switch of Quiz 10. %first event is an arrival timenow=0.
66
..1) ]. time=[time(b4arrival) arrival time(˜b4arrival)].admits. b4depart=time<depart. % clear current event if (eventnow==1) % arrival arrival=timenow+exponentialrv(lam. end elseif (eventnow==-1) %departure n=n-1. n=0. while (timenow<tmax) M((timenow<=t)&(t<time(1)))=n. blocks=0. %one more block.

1 RY (τ ) = e−|τ | 2 Quiz 11.5(1 + −1) = 0
(1)
The autocorrelation of the output is
1 1
RY [n] =
i=0 j=0
h i h j R X [n + i − j] 1 n=0 0 otherwise
(2) (3)
= 2R X [n] − R X [n − 1] − R X [n + 1] =
2 Since µY = 0.Quiz Solutions – Chapter 11
Quiz 11.2.2 The expected value of the output is
∞ ∞ −τ
h(u)h(τ + u) du =
∞ −τ
1 e−u e−τ −u du = eτ 2
(4)
(5)
µY = µ X
n=−∞
h n = 0. we have RY (τ ) =
0 ∞
e−u e−τ −u du = e−τ
0
∞
1 e−2u du = e−τ 2
(3)
For τ < 0. RY (τ ) = Hence.1 By Theorem 11. For τ < 0. The variance of Yn is Var[Yn ] = E[Yn ] = RY [0] = 1.
67
. µY = µ X
∞ −∞
h(t)dt = 2
0
∞
e−t dt = 2
(1)
Since R X (τ ) = δ(τ ). we 2 can double check. Just to be safe though. we can deduce that RY (τ ) = 1 e−|τ | by symmetry. the autocorrelation function of the output is RY (τ ) =
∞ −∞ ∞
h(u)
−∞
h(v)δ(τ + u − v) dv du =
∞ −∞
h(u)h(τ + u) du
(2)
For τ > 0.

6 SX(f) 0. RX = I. Moreover.5.1 0.8. each Yn has expected value E[Yn ] = µ X ∞ n=−∞ h n = 0. following Theorem 11.5 to ﬁnd the autocorrelation function
∞ ∞
RY [n] =
i=−∞ j=−∞
h i h j R X [n + i − j]. we need to ﬁnd the covariance matrix CY . which equals the correlation matrix RY since Y has zero expected value.
(1)
Despite the fact that R X [k] is an impulse. using Equation (1) is surprisingly tedious because we still need to sum over all i and j such that n + i − j = 0. it is simpler to observe that Y = HX where X = X 30 X 31 X 32 X 33 X 34 X 35 and ⎡ ⎤ 1 1 1 1 0 0 1 H = ⎣0 1 1 1 1 0⎦ .3 By Theorem 11. Fo ﬁnd the PDF of the Gaussian vector Y.13 with µX = 0 and A = H. Y = Y33 Y34 Y35 is a Gaussian random vector since X n is a Gaussian random process.x 10 8 0.4 0.2 −0.5.2 0 −15 −10 −5 0 f 5 10 15 SX(f) 6 4 2 0 −1500−1000 −500 10 R (τ) 5 0 −5 −2 −1 0 τ 1 x 10 2
−3
0 f
500 1000 1500
10 RX(τ) 5 0 −5 −0. In this problem. Thus E[Y] = 0.7.2
X
(a) W = 10
(b) W = 1000
Figure 6: The autocorrelation R X (τ ) and power spectral density S X ( f ) for process X (t) in Quiz 11. or by directly applying Theorem 5. One way to ﬁnd the RY is to observe that RY has the Toeplitz structure of Theorem 11. Quiz 11. 4 0 0 1 1 1 1 (2)
(3)
In this case. 68
.6 and to use Theorem 11. Since R X [n] = δn .1 0 τ 0. we obtain RY = HRX H . by Theorem 11. the identity matrix.

Y Quiz 11. X n+1 = 400 400 (4)
to ﬁnd the mean square error.9 h = R−1 RXn X n+1 = Xn 0. X n+1 = Xn 0. CY = RY = HH = 16 2 3 4
(4)
It follows (very quickly if you use M ATLAB for 3 × 3 matrix inversion) that ⎡ ⎤ 7/12 −1/2 1/12 1 −1/2⎦ .9 R X [0] R X [1] = 0. one approach is to follow the method of Example 11. the PDF of Y is f Y (y) = 1 (2π )3/2 [det (CY )]1/2 1 exp − y C−1 y . C−1 = 16 ⎣−1/2 Y 1/12 −1/2 7/12 Thus.81 81 = . In this case.4 This quiz is solved using Theorem 11.1 R X [1] R X [0] 0. (7)
Equation (7) shows that one of the nicest features of the multivariate Gaussian distribution is that y C−1 y is a very concise representation of the cross-terms in the exponent of f Y (y).9 1.Thus
⎡ ⎤ 4 3 2 1 ⎣ 3 4 3⎦ .1 0.9 R X [1]
−1
(1)
(2)
The MMSE linear ﬁrst order ﬁlter for predicting X n+1 at time n is the ﬁlter h such that ← − 1.1 0.81 X n−1 R X [2] = . Xn = X n−1 X n and RXn = and RXn X n+1 = E 1.9 for the case of k = 1 and M = 2.13 and to directly calculate ˆ (5) e∗ = E (X n+1 − X n+1 )2 .9 400 261 (3)
It follows that the ﬁlter is h = 261/400 81/400 and the MMSE linear predictor is 81 261 ˆ X n−1 + Xn. Y 2
(5)
(6)
A disagreeable amount of algebra will show det(CY ) = 3/1024 and that the PDF can be “simpliﬁed” to 16 7 2 7 2 1 2 y33 + y34 + y35 − y33 y34 + y33 y35 − y34 y35 exp −8 f Y (y) = √ 3 12 12 6 6π .9 1. 0. L 69
.1 1 0.

e∗ = E L ← − X n+1 − h Xn
2
(6) (7) (8)
← − ← − = E (X n+1 − h Xn )(X n+1 − h Xn ) ← − ← − = E (X n+1 − h Xn )(X n+1 − Xn h ) After a bit of algebra.9 400 1451 recalling that the blind estimate would yield a mean square error of Var[X ] = 1. the mean square error is 1 506 ← − 0.81 81 261 e∗ = R X [0] − h RXn X n+1 = 1. we obtain ← − ← − ← − e∗ = R X [0] − 2 h RXn X n+1 + h RXn h L
(9) (10)
← − with the substitution h = R−1 RXn X n+1 . the average power of X (t) is E X 2 (t) =
∞ −∞ W −W
SX ( f ) d f =
5 d f = 10 Watts W
(1)
(2) The autocorrelation function is the inverse Fourier transform of S X ( f ).13(b). In any case. we obtain Xn e∗ = R X [0] − RXn X n+1 R−1 RXn X n+1 L Xn ← − = R X [0] − h RXn X n+1 (11) (12)
Note that this is essentially the same result as Theorem 9. Consulting Table 11. graphs of S X ( f ) and R X (τ ) appear in Figure 6. (13) L 0. Quiz 11.3487. It is noteworthy that the result is derived in a much simpler way in the proof of Theorem 9. we can derive the mean square error for an arbitary prediction ← − ˆ ﬁlter h. Since X n+1 = h Xn .1 − = = 0.7 by using the orthoginality property of the LMSE estimator.This method is workable for this simple problem but becomes increasingly tedious for higher order ﬁlters. Instead.1. 70
. we note that 1 f S X ( f ) = 10 rect (2) 2W 2W It follows that the inverse transform of S X ( f ) is sin(2π W τ ) R X (τ ) = 10 sinc(2W τ ) = 10 (3) 2π W τ (3) For W = 10 Hz and W = 1 kHZ.7 with Y = Xn . X = X n+1 and ← − ˆ a = h . we see that observing X n−1 and X n improves the accuracy of our prediction of X n+1 .1.5 (1) By Theorem 11.

we recall the property that g(τ − τ0 ) has Fourier transform G( f )e− j2π f τ0 . S X Y ( f ) = H ( f )S X ( f ) = (2) Again by Theorem 11. Thus the Fourier transform of R X Y (τ ) = R X (τ − t0 ) = g(τ − t0 ) is S X Y ( f ) = S X ( f )e− j2π f t0 . That is.7 Since Y (t) = X (t − t0 ). From Table 11. a0 Consulting with the Fourier transforms in Table 11.1. τ ) = E [X (t)Y (t + τ )] = E [X (t)X (t + τ − t0 )] = R X (τ − t0 ) (1)
We see that R X Y (t. SY ( f ) = H ∗ ( f )S X Y ( f ) = |H ( f )|2 S X ( f ).Quiz 11. then
∞
S X (φ) =
n=−∞
10δ[n]e− j2π φn = 10
(1)
Thus. the discrete time impulse δ[n] has a ﬂat discrete Fourier transform. we see that
2 2a0 1 2a0 SX ( f ) = = 2 2 + (2π f )2 a0 a0 a0 + (2π f )2
(2)
The RC ﬁlter has impulse response h(t) = a1 e−a1 t u(t). 2 [a1 + j2π f ] a0 + (2π f )2 (4) a1 a1 + j2π f (3)
. From Table 11. τ ) = R X Y (τ ) = R X (τ − t0 ).8 We solve this quiz using Theorem 11. (This quiz is really lame!) Quiz 11.1.6 In a sampled system. R X Y (t. if R X [n] = 10δ[n]. (2)
Quiz 11.1.17.17. First we need some preliminary facts. Let a0 = 5.17. H( f ) = (1) Theorem 11. R X [n] = 10δ[n].000 so that 1 (1) R X (τ ) = a0 e−a0 |τ | . where u(t) is the unit step function and a1 = 1/RC where RC = 10−4 is the ﬁlter time constant. 71 (5) 2a0 a1 .

Using partial fractions and the Fourier transform table. the latter method is actually less algebra. the output signal has almost as much power as the input. 2
K1 =
.
(12)
The average power of the Y (t) process is RY (0) = a1 2 = .
(9)
(10)
Consulting with Table 11.000 rad/sec. we obtain RY (τ ) =
2 a1 e−a0 |τ | − a0 a1 e−a1 |τ | 2 2 a1 − a0
. SY ( f ) = 2 2 2a0 a0 + (2π f )2 2a1 a1 + (2π f )2 2 a0
K0 K1 + + (2π f )2 a1 + (2π f )2
2 −2a0 a1 2 2 a1 − a0
(8)
2 2a0 a1
2 a1 − a0
. we see that RY (τ ) = K0 K1 a e−a0 |τ | + 2 a1 e−a1 |τ | 2 0 2a0 2a1 (11)
Substituting the values of K 0 and K 1 . Since the RC ﬁlter has a 3dB bandwidth of 10.Note that |H ( f )|2 = H ( f )H ∗ ( f ) = Thus.000 rad/sec and the signal X (t) has most of its its signal energy below 5.
2 2 2a0 2a1 K0 K1 + 2 . a1 + a0 3 (13)
Note that the input signal has average power R X (0) = 1. we can either use basic calculus and ∞ calculate −∞ SY ( f ) d f directly or we can ﬁnd RY (τ ) as an inverse transform of SY ( f ). In particular. some algebra will show that SY ( f ) = where K0 = Thus. SY ( f ) = |H ( f )|2 S X ( f ) =
2 2a0 a1 2 2 a1 + (2π f )2 a0 + (2π f )2 2 a1 a1 a1 = 2 (a1 + j2π f ) (a1 − j2π f ) a1 + (2π f )2
(6)
(7)
(3) To ﬁnd the average power at the ﬁlter output.
72
.1.

SY X ( f ) = S X ( f ). 4 10 104 (4)
The noise power spectral density can be written as S N ( f ) = N0 rect f 2B = 1 f rect 2B 2B . the optimal ﬁlter is ˆ H( f ) = SX ( f ) = SX ( f ) + SN ( f )
1 104 1 104
rect +
f 104 1 2B
rect
f 104
rect
f 2B
.9 This quiz implements an example of Equations (11. R N (0) = Var[N ] = 1.
(6)
73
. Comment: Since the text omitted the derivations of Equations (11. it follows that SY ( f ) = S X ( f ) + S N ( f ).146) and (11. (2) Since R X (τ ) = sinc(2W τ ).147). we see from Table 11. we note that Example 10. where W = 5.24 showed that RY (τ ) = R X (τ ) + R N (τ ).147) for a system in which we ﬁlter Y (t) = X (t) + N (t) to produce an optimal linear estimate of X (t). at peace with the derivations. (2) RY X (τ ) = R X (τ ).147).146).000 Hz. This implies R N (0) =
∞ −∞
SN ( f ) d f =
B −B
N0 d f = 2N0 B
(3)
Thus N0 = 1/(2B). decreasing the single-sided bandwidth B increases the power spectral density of the noise over frequencies | f | < B.Quiz 11.146) and to calculate the mean square error e L ∗ using Equation (11.1 that SX ( f ) = 1 f rect .146) and (11. Because the noise process N (t) has constant power R N (0) = 1. Taking Fourier transforms. (5)
From Equation (11. The ˆ solution to this quiz is just to ﬁnd the ﬁlter H ( f ) using Equation (11. (1)
Now we can go on to the quiz. (1) Since µ N = 0.

000.000 B
(9)
To obtain MSE e∗ ≤ 0. Finally. S N ( f ) = 1/2B over frequencies | f | < W .147).05.10 It is fairly straightforward to ﬁnd S X (φ) and SY (φ). we note that we can choose B very large and also achieve MSE e∗ = 0. We can go back and consider the case B > W later. Thus increasing B spreads the constant 1 watt of power of N (t) over more bandwidth.ˆ ˆ (3) We produce the output X (t) by passing the noisy signal Y (t) through the ﬁlter H ( f ).5 × 104 guarantees e∗ ≤ 0. The following M ATLAB program generates and plots the functions shown in Figure 8 74
. we need to whether B ≤ W . 1 ˆ + 1 (10) H( f ) = 104 2B ⎩ 0 otherwise. when B > W = 5000. ˆ the Wiener ﬁlter H ( f ) is an ideal (ﬂat) lowpass ﬁlter ⎧ 1 ⎨ 104 | f | < 5. The mean square error is e∗ L =
1 1 104 2B 1 1 −5000 104 + 2B 5000
df =
1 2B 1 104
+
1 2B
=
1
B 5000
+1
(11)
In this case. As B shrinks. for all values of B. let’s suppose B ≤ W . L Although this completes the solution to the quiz. The result is that the MSE goes down. B ≥ 9. Since the problem asks us to L ﬁnd the largest possible B. In L particular.05 requires B ≤ 5. When B ≤ W . what is happening may not be obvious. As B is decreased.
(8)
To evaluate the MSE e∗ . The noise power is always Var[N ] = 1 Watt.16 Hz. The Wiener ﬁlter removes the noise that is outside the band of the desired signal. the MSE is e∗ L =
1 1 104 2B 1 1 −B 104 + 2B B
df =
1 104 1 104
+
1 2B
=
1 1+
5. In this case.05. the mean square error of the estimate is e∗ = L =
∞ −∞ ∞ −∞
S X ( f )S N ( f ) df SX ( f ) + SN ( f )
1 104 1 104
(7)
f 2B f 2B
rect
f 104 f 104
1 2B
rect rect
rect
+
1 2B
d f.000/19 = 263. Two examples of the ﬁlter H ( f ) are shown in Figure 7. the ﬁlter suppresses less of the signal of X (t). Thus as ˆ B descreases. L Quiz 11. From Equation (11. but only over a bandwidth B that is decreasing. the PSD S N ( f ) becomes increasingly tall. The only thing to keep in mind is to use fftc to transform the autocorrelation R X [ f ] into the power spectral density S X (φ). the ﬁlter H ( f ) makes an increasingly deep and narrow notch at frequencies ˆ | f | ≤ B.

%impulse/filter response: M=10 SY10=sx.1 H(f) 0.m N=32.5*[1 1]. H2=fft(h2.* ((abs(H2)). when M = 10.26.abs(SY10)). stem(0:N-1. Although these imaginary parts have no computational signiﬁcance.ylabel(’S_X(n/N)’). the ﬁlter H (φ) ﬁlters out almost all of the high frequency components of X (t). %PSD of Y for M=2 xlabel(’n’).ˆ2).1*ones(1.N). %impulse/filter response: M=2 SY2=SX. SX=fftc(rx.ylabel(’S_{Y_{10}}(n/N)’).N). h10=0. However.
Relative to M = 2. %autocorrelation and PSD stem(0:N-1. figure. they tend to confuse the stem function.abs(sx)). h2=0. the low pass moving average ﬁlter for M = 10 removes the high frquency components and results in a ﬁlter output that varies very slowly. xlabel(’n’). xlabel(’n’).10). As an aside.
75
. H10=fft(h10.9.*((abs(H10)).ˆ2). SY2 and SY10 in mquiz11 should all be realvalued vectors.ylabel(’S_{Y_2}(n/N)’).5 0 H(f) −5000 −2000 0 f 2000 5000
1 0.abs(SY2)). In the context of Example 11.5 0
−5000
−2000
0 f
2000
5000
B = 500
B = 2500
Figure 7: Wiener ﬁlter for Quiz 11.
%mquiz11. figure. rx=[2 4 2]. note that the vectors SX. we generate stem plots of the magnitude of each power spectral density.N). Hence. stem(0:N-1. the ﬁnite numerical precision of M ATLAB results in tiny imaginary parts.

SY (n/N ) for M = 2.10.10 SX(n/N)
5
0
0
5
10
15 n
20
25
30
35
10 SY (n/N)
2
5
0
0
5
10
15 n
20
25
30
35
10 SY (n/N)
10
5
0
0
5
10
15 n
20
25
30
35
Figure 8: For Quiz 11. and Sφ (n/N ) for M = 10 using an N = 32 point DFT. graphs of S X (φ).
76
.

2 −0.5 1
(3)
where si .6 0.6 0.9
P X n+1 = 0|X n = 1 = 0.2 0 0 ⎦ Pn = S−1 Dn S = ⎣0.01 0.4 0.4)n ⎣ 0 (4) −0. the Markov chain and the transition matrix are ⎡ ⎤ 0.99 0.4 0. we can conclude that P X n+1 = 1|X n = 0 = 0.4 0.2 0.6 0 0.4 λ3 = 1
(1)
(2)
We can diagonalize P into ⎤⎡ ⎡ ⎤ ⎤⎡ −0.2 0.1
(2)
These conditional probabilities correspond to the transition matrix and Markov chain:
0.5 1 −0.6 −0.5 1 λ1 0 0 0 −1 ⎦ 0 1 ⎦ ⎣ 0 λ2 0 ⎦ ⎣ 1 P = S−1 DS = ⎣ 0.6 P = ⎣0. is the left eigenvector of P satisfying si P = λi si . the ith row of S.5 0 −0.6 0. we are given the conditional probabilities P X n+1 = 0|X n = 0 = 0.2 The eigenvalues of P are λ1 = 0 λ2 = 0.1 The system has two states depending on whether the previous packet was received in error.2 0.01
0.2⎦ + (0.6 0.6 0.5 0.99 P X n+1 = 1|X n = 1 = 0.10 0. From the problem statement.2⎦ 1 0 1 0 0.1
1
P=
0.6 0.5 0.Quiz Solutions – Chapter 12
Quiz 12.90
(3)
Quiz 12.01
0
0.2 From the problem statement.99 0.4 0.9 (1)
Since each X n must be either 0 or 1. Algebra will verify that the n-step transition matrix is ⎡ ⎡ ⎤ ⎤ 0.4 0 0 λ3 0.6 0.6 0.5 0 0.2 0.2 0.5 −0.2 0.2 Quiz 12.6 0.3 The Markov chain describing the factory status and the corresponding state transition matrix are 77
.

9 0. (3) (2)
The states in C1 and C3 are aperiodic. 5.n = P [K > n|K > n − 1] = Pn−1. the class C1 is never left. Similarly.1 + 0.9 0. Once the system exits C2 . 2. the states in C2 are never reentered.4 The communicating classes are C1 = {0. 1.1π0 and π2 = π1 .1 0 0 1⎦ P=⎣ 0 1 0 0
(1)
2
With π = π0 π1 π2 . .
1
…
78
.1) = 1 It follows that the limiting state probabilities are π0 = 5/6. C1 is a recurrent class.. π2 = 1/12. The states in C2 have period 2.0 P [K > n] P [K > n − 1] P [K = n] = P [K = n|K > n − 1] = P [K > n − 1] (1) (2) (3) The Markov chain resembles
P[K=5] P[K=4] P[K= 1] P[K=2] P[K=3]
0
1
1
1
2
1
3
1
4
. Once the system enters a state in C1 . . 1} C2 = {2. On the other hand. the system of equations π = π P yields π1 = 0. 3} C3 = {4.5 At any time t.0.1
0
1
1
1
⎡ ⎤ 0.. Quiz 12. the state n can take on the values 0. 6} (1) π1 = 1/12. the states in C3 are recurrent.. Thus the states in C1 are recurrent. The state transition probabilities are Pn−1. That is. the states in C2 are transient. This implies π0 + π1 + π2 = π0 (1 + 0. Quiz 12. .

Equation (5) implies π2 = π1 − π0 P [K = 2] = π0 (P [K > 1] − P [K = 2]) = π0 P [K > 2] (8) (7) (4) (5) k = 1. . we obtain π1 = π0 (1 − P [K = 1]) = π0 P [K > 1] Similarly. From Problem 2. When we apply we recall that
∞ k=0 πk ∞ k=0 P[K
(9)
= 1. Thus the period of state 0 is d = 2. When the counter expires. From Equation (4). including state 0. . The system state is the time until the counter expires.5. . (2) To ﬁnd the stationary probabilities. . πk−1 = π0 P [K = k] + πk . the number of transitions need to return to state 0 is always a multiple of 2. This implies πn = P [K > n] E [K ] (10)
This Markov chain models repeated random countdowns.The stationary probabilities satisfy π0 = π0 P [K = 1] + π1 . n=0 > k] = E[K ]. the system is in state 0. . π1 = π0 P [K = 2] + π2 . Since we spend one unit of time in each state.11. then W has a discrete PMF representing the remaining time of the counter at a time in the distant future. . we have k − 1 units of time left after the state 0 counter reset. Quiz 12. If we have a random variable W such that the PMF of W satisﬁes PW (n) = πn . we solve the system of equations π = πP and 3 i=0 πi = 1: π0 = (3/4)π1 + (1/4)π3 π1 = (1/4)π0 + (1/4)π2 π2 = (1/4)π1 + (3/4)π3 1 = π0 + π1 + π2 + π3 79 (1) (2) (3) (4)
. 2.
(6)
This suggests that πk = π0 P[K > k]. we obtain π0 ∞ P[K > k] = 1. and we randomly reset the counter to a new value K = k and then we count down k units of time. We verify this pattern by showing that πk = π0 P[K > k] satisﬁes Equation (6): π0 P [K > k − 1] = π0 P [K = k] + π0 P [K > k] .6 (1) By inspection.

Lastly. we choose π0 so the state probabilities sum to 1: 16 2 5 1 = π0 + π1 + π2 + π3 = π0 1 + + + 2 = π0 (7) 3 3 3 It follows that the state probabilities are π0 = 3 16 π1 = 2 16 π2 = 5 16 π3 = 6 16 (8)
(3) Since the system starts in state 0 at time 0.7 The Markov chain has the same structure as that in Example 12. To determine whether state 0 is recurrent.(3/4) 1 .
(1)
Thus the CDF of T00 satisﬁes FT00 (n) = 1− P[T00 > n] = 1−1/n α .(2/3) a 1 .(1/2)
a
1
1 . The only difference is the modiﬁed transition rates:
1 (1/2)a (2/3)a (3/4)
a
(4/5)
a
0
1.Solving the second and third equations for π2 and π3 yields π2 = 4π1 − π0 π3 = (4/3)π2 − (1/3)π1 = 5π1 − (4/3)π0 (5)
Substituting π3 back into the ﬁrst equation yields π0 = (3/4)π1 + (1/4)π3 = (3/4)π1 + (5/4)π1 − (1/3)π0 (6)
This implies π1 = (2/3)π0 .22.14 to ﬁnd the limiting probability that the system is in state 0 at time nd: lim P00 (nd) = dπ0 = 3 8 (9)
n→∞
Quiz 12. It follows from the ﬁrst and second equations that π2 = (5/3)π0 and π3 = 2π0 . which occurs with probability P [T00 1 > n] = 1 × 2
α
2 × 3
α
n−1 × ··· × n
α
=
1 n
α
.(4/5)a
a
2
3
4
…
The event T00 > n occurs if the system reaches state n before returning to state 0. we can use Theorem 12. nα
(2)
80
. we observe that for all α > 0 P [V00 ] = lim FT00 (n) = lim 1 −
n→∞ n→∞
1 = 1.

) To determine whether the chain is null recurrent or positive recurrent. the Markov chain is positive recurrent. Applying this result.24. ( We also note that if α = 0. Quiz 12. In this problem. On the other hand. it will be simpler to use the result of Problem 2. we did this by deriving the PMF PT00 (n).11 which says that ∞ P[K > k] = k=0 E[K ] for any non-negative integer-valued random variable K . we need to calculate E[T00 ]. 1/n α ≥ 1/n and it follows that
∞
E [T00 ] ≥ 1 +
n=1
1 = ∞.5. Since the chain has only one communicating class. n
(4)
We conclude that the Markov chain is null recurrent for 0 < α ≤ 1. In Example 12. the expected time to return to state 0 is
∞ ∞
E [T00 ] =
n=0
P [T00 > n] = 1 +
n=1
1 . ∞ 1 E [T00 ] = 2 + .Thus state 0 is recurrent for all α > 0. then all states are transient.8 The number of customers in the ”friendly” store is given by the Markov chain
(1-p)(1-q) p (1-p)(1-q) p (1-p)(1-q) p (1-p)(1-q)
0
(1-p)q
1
(1-p)q
×××
i
(1-p)q (1-p)q
i+1
×××
81
. for α > 1. (5) nα
n=2
Note that for all n ≥ 2
1 ≤ nα
∞
n n−1
dx xα
(6)
This implies E [T00 ] ≤ 2 + =2+
n
n=2 n−1 ∞ dx 1
dx xα
(7) (8)
xα
x −α+1 =2+ −α + 1
∞
=2+
1
1 <∞ α−1
(9)
Thus for all α > 1. all states are recurrent. nα
(3)
For 0 < α ≤ 1.

.01 p3 = 2 p2 + 3 p4 5. for α ≥ 1 or. . i = 0. we have that πi = π0 α i where p .In the above chain. p3 in terms of p2 and so on. (1 − p)q (1) (2)
Since Equation (2) holds for i = 0.13 with state space partitioned between S = {0.01 p2 = 2 p1 + 3 p3 3. 1. 5. This implies πi+1 = p πi .}.1 since the task completes at rate 3 per msec and the processor reboots at rate 0. α= (1 − p)q Requiring the state probabilities to sum to 1.01
1
3 0. we obtain the following useful equations for the stationary distribution. an existing customer gets one unit of service and then departs the store. i + 2. the limiting state probabilities do not exist. . 381 (1)
. . .01 p4 = 2 p3
We can solve these equations by working backward and solving for p4 in terms of p3 .01 0. equivalently.9 The continuous time Markov chain describing the processor is
2 2 2 2
0
3. Quiz 12. πi p = πi+1 (1 − p)q. . 620 p0 1.
∞ ∞
(3)
πi = π0
i=0 i=0
αi =
π0 = 1.01 0. the limiting state probabilities are πi = (1 − α)α i . 1. . we have that for α < 1. By applying Theorem 12. . i} and S = {i + 1. yielding p4 = 20 p3 31 p3 = 620 p2 981 p2 = 82 19620 p1 31431 p1 = 628. 014. we note that (1 − p)q is the probability that no new customer arrives. . From the Markov chain. (5)
In addition. . .01 p1 = 2 p0 + 3 p2 5. we see that for any state i ≥ 0. . . 2. p ≥ q/(1 − q). 1−α
(4)
Thus for α < 1. . 1.01
2
3
3
3
4
Note that q10 = 3.1 per msec and the rate to state 0 is the sum of those two rates.

401 and the stationary probabilities are p0 = 0. 2. . 443. the stationary probabilities must satisfy pn = (ρ/n) pn−1 n = 1.10 The M/M/c/∞ queue has Markov chain
λ λ λ λ λ
(2)
0
µ
1
2µ cµ
c
cµ
c+1
cµ
From the Markov chain. .1606 p3 = 0. c (ρ/c) pn−1 n = c + 1. 2. . . pn = 1 yields
c
(2)
p0 =
n=0
ρ c ρ/c ρ /n! + c! 1 − ρ/c
n
−1
(3)
83
. c + 2. c n−c c p0 (ρ/c) ρ /c! n = c + 1. . . . .Applying p0 + p1 + p2 + p3 + p4 = 1 yields p0 = 1.1015 p4 = 0.0655 Quiz 12. c + 2. . . 014.2573 p2 = 0. . 381/2.4151 p1 = 0. (1)
It is straightforward to show that this implies pn = The requirement that
∞ n=0
p0 ρ n /n! n = 1. . . .