Académique Documents
Professionnel Documents
Culture Documents
Introduction to Probability
with Statistical Applications
Geza Schay
1.1.1.
a) The sample points are
b) The event that corresponds to the statement at least one tail is obtained is
c) The event that corresponds to at most one tail is obtained is
1.1.3.
a) Four different sample spaces to describe three tosses of a coin are:
an even # of
s, an odd # of
c) It is not possible to nd an event corresponding to the statement at most one tail is obtained
in three tosses in every conceivable sample space for the tossing of three coins, because some
sample spaces are too coarse, that is, the sample points that contain this outcome also contain
opposite outcomes. For instance, in
above, the sample point an even # of s contains
for which our statement is true and the outcome
the outcomes
for which it is not true.
1.1.5.
In the 52-element sample space for the drawing of a card
The card drawn is neither red, nor odd, nor a face card are
!" # $ $ $ $ !" $ %
and
, and
b) statements corresponding to the events
& ' () * + * , * - * %
' (# !" # $ $ $ $ !" $ %
'
, and .
are /
0
1
' 0
The Ace of hearts or a heart face card is drawn, and 2
An even numbered black card is
1
drawn.
1.1.7.
Three possible sample spaces are:
%
$ 3 ' (
The 365 days of the year
1
$
$
'
'
%
1.2.1.
( !
a)
b)
c)
d)
e)
1.2.3.
,
or
1.2.5.
!
"
"
$ #
"
$ !
"
"
"
"
"
"
"
$ #
'
"
$ !
$ #
'
"
'
$ !
"
$ !
'
"
"
'
&
'
&
"
"
"
&
'
'
'
"
"
$ #
"
$ !
'
'
"
$ !
'
"
$ !
1.2.7.
a)
but
b)
and
c)
and
1.2.9.
The Venn diagram below illustrates the relation
the diagram, we have
and
Similarly,
!
#
"
Figure 1.
&
"
1.2.11.
that is, that whenever
then
Then
1. Assume that
or
or
On the other hand, clearly
Thus,
implies
2. Conversely, assume that
that is, that
or
Hence, if
then must also belong to
which means that
Alternatively, by the de nition of unions,
and so, if
then substituting
for
in the previous relation, we obtain that
implies
1.3.1.
a) The event corresponding to
is 4 or 5 is the shaded region consisting of the fourth
and fth columns in the gure below, that is,
and
.
!
'
'
'
'
'
'
'
'
Figure 2.
&
%
'
or
$
&
is 4 or 5
corresponding to
& %
Figure 3.
'
corresponding to or
is
&
Figure 4.
"
"
% $
and
%
"
"
"
% $
is
%
"
Figure 5.
$
$ &
and
%
$
%
$
is
%
"
"
"
Figure 6.
"
corresponding to
1.3.3.
1.3.5.
or (that is, at least one of them) is certain to occur.
2.1.1.
Let
2.1.3.
If
then and have no common element. Hence
common element either.
Alternatively, if
then
The proofs of the other cases just need changing letters.
2.1.5.
and
2.2.1.
a)
b)
Figure 7.
c)
2.2.3.
2.2.5.
a)
2.2.7.
a)
2.3.1.
b)
2.3.3.
b)
2.3.5.
b)
2.3.9.
a)
2.3.11.
a)
b)
c)
d)
2.4.1.
,
,
1
1
1
10
35
70
126
15
35
56
84
20
21
28
36
6
10
15
7
8
4
5
1
1
1
1
1
21
56
126
1
7
28
84
8
36
2.4.3.
2.4.5.
2.4.7.
a)
b)
for any
2.4.9.
a)
b)
2.5.1.
a) ,
b)
,
,
c)
d)
2.5.3.
,
a)
b)
,
c)
,
d)
2.5.5.
a)
b)
2.5.7.
indistinguishable balls into
distinguishable boxes. It
a) This is like putting
can be done in
ways.
b) There are 9 spaces between the 10 balls if we put them in a row. With two dividing
bars, we can divide the balls into 3 groups. So, the number of ways of dividing them into 3
nonempty groups is
2.5.9.
ways.
You have to choose boxes out of This can be done in
"
&
'
$ (
* $
'
F GN
- . / 0
1
2 3
4
- 5 .
1
6
; <
; <
F GH
>
<
>
F GK
N I
F QL I
F GL
F GM
P Q
F L
UQ S T
Z U
V ]
Z U
S V
b
^ S V ^ S\ ^
b \c
X Y
[ S TU S _
[ ^
c d
` a
X Y
X `
~
h f g
g ij g
l m n
oh g
p n
hq g
g
p n
s n n
p n
t v
s n n
m l n
x y
w
u {
{ |
~
{ |
3.1.1.
a) P(
b) P(
c) P(
d) P(
e) P(
f) P(
g) P(
h) P
3.1.3.
,
,
,
,
,
,
P(
and
Thus, by Axiom 3, P
P
P
Similarly,
P
P(
P
Adding, we get P
P
P
2P
P(
Here, again by Axiom 3, we have P
P
P(
P(
Hence, P
P
P(
P
and so P
P
P
P
3.1.5.
we have, as in Problem 3.1.3, P
P
P
and so
Since
we always have P
P
P
Thus, P
P
P
if and only
if P
P
This relation is true, in particular, if
that is, if
But
!
P
P
can also hold if
but P(
because P
P(
P
for any and
3.1.7.
a) This result follows at once from Theorem 3.1.2 because we are subtracting the (by Axiom
from P
P
on the right of Equation 3.1.1 to get P(
1) nonnegative quantity P
).
b) Apply the result of Part a) with
in place of
and in place of
Then we get
"
P
P
P
Now, apply the result of Part a) to
and we
"
obtain P
P
P
P
Since unions are associative, this proves the
required result.
$
c) This relation can be proved by induction: As seen above, it is true for #
and 3. For
any larger # assume the formula to be true for #
Then we can prove it for # as follows:
5 6
A B
and by Part a), P 7 8 9< =: ; < ?
P( C DG HE F G J K P L M N J O
P % &' ( ) * ' + , P - % &0 1. / 0 3
N Q R
/ 2
2 4
;
>
>
@
F I
S
V
W
X
Y
Z
[
\]
By the induction hypothesis, P L P S T R U
P ^ _ ` a b and so, putting all these relations
[
together, we get P c d e` f g _ ` a h i jk l m P n o p q r
3.2.1.
a) s t u o s o v w o s x s w o s x v w o v o s w o v x s w o v x v w x s o s w x s o v w x s x v w
x v o s wx v o v wx v x s y r
10
b) P(o and x q t
t
r
c) There are 6 possible unordered pairs, 4 of which are favorable. So, P(o and x q t
t
r
d) Here we are drawing without replacement and so each pair consists of two different
cards. Thus, each unordered pair corresponds to two ordered pairs and therefore each one has
t
r In Example 3.2.2, some unordered pairs correspond to two ordered
probability
pairs and some to one.
3.2.3.
We did not get P(at least one six) = 1, in spite of the fact that on each throw the probability of
getting a six is , and 6 times is 1, for two reasons: First, we would be justi ed in taking
the six times here only if the events of getting a six on the different throws were mutually
exclusive then the probability of getting a six on one of the throws could be computed by
Axiom 3 as
but these are not mutually exclusive events. Second, the event of getting at
least one six is not the same as the event of getting a six on the rst throw, or on the second,
or etc.
3.2.5.
P(different numbers with three dice
3.2.7.
people can be seated in
ways. The number of favorable cases is
because the group of men can start at any one of the
seats and must be followed by the
group of women, and in each case the men can be permuted
ways amongst themselves
and the women ways. Thus, P
3.2.9.
This problem is like sampling good and bad items without replacement. The good items are
the player s numbers and the bad ones are the rest. Thus,
! & "
P(jackpot
P QR P Q
S T Q
. /
$ !
#
%#
&
$ %
'
"
%#
Z
c
A B
) $ % *
and P(match 5
: ; 8 7 :9 ; < =
X W ] X_ X` W \ a b
G H
?
4 75 68 4 5
P UR P
V WX Y Z W[ \ ]
"
,
-
. 0 /
1
,
. 2 3
f g
4
i
>
G J H
I E
K
G L M
H
E
F
@
N
3.2.11.
a)
b)
c)
3.2.13.
To get 5 cards of different denominations, we may rst choose the 5 denominations out
of the 13 possible ones and then choose one card from the 4 cards of each of the selected
l
j kl m
j kl m
j kn m
j n
m k
lk r
l r
k
p
k
j ls r
iu u
i v w x
ls r
straights and
~
z | } { z {
~
|
z ~ {
11
ve cards of the same suit, which are very valuable hands, while the other cases of different
denominations are poor hands.)
3.2.15.
For the pair we have 13 possible denominations and then for the triple, 12 possible denominations. For the pair we have
choices from the 4 cards of the selected denomination and
3.2.17.
In poker dice, we have 6 possible numbers for the pair and then 5 for the triple. These
ways. Thus
selections can be ordered in
"
# $
&
' () * + , (
%
. /
5 6
5 6
5 6
5 6
2 8
2 8
<
<
<
<
<
/ (
3 0
2 3
9: 7
9 : 7
9 : 7
3 0
/ 8
3 0
/ 8 3
9: 7
B <
<
<
<
>
<
<
<
@
8
7 D
2 3
/ 8
B <
<
B <
<
<
<
<
B <
<
<
<
K M
KL
<
<
<
/ (
2 (
<
<
<
<
FG
JI
K L
KL
/ 8
3 0
>
2 8
2 8
) 3 .
/ 3
) 3 .
/ 3
) 3 .
) 3 .
<
/ 3
2 3
KL
K M
JI
K P
QF R
SI
T K U V W N X
KL
Q IR
KL
K P
H
Y
KL
K P
KL
Q IR
KL
K M
KL
]^
ZL
K M
N _
KL
ZM
ZM
KL
KL
ZM
KL
K L
KM
ZM
ZL
12
ZM
KL
]^
KL
N _
ZM
3.3.7.
K N
K N
Z JI [
V
K ^ N
Z JI [
V
K N
^
Z JI [
V
K U N
^
Z JI [
V
K N
Z JI [
Z JI [
O
Figure 8.
3.3.9.
Similarly, the probability is the
The probability that a ball picked at random is red is
same for white and also for blue. Thus, the probability for any color combination in a given
order for six independently chosen balls is
We can obtain two of each color in
H
QI
Z QI [
Z J
13
under 30 <
P(under 30)
P(under 30)
P(
under 30 <
C
3.4.7.
Whether the selected girl is the rst, second or third child in the family, her siblings, in the
order of their births, can be or In two of these cases does the family have two
girls and one boy. Thus, P(two girls and one boy one child is a girl
3.4.9.
The number of ways of drawing two Kings, which are also face cards, without replacement,
is and the number of ways of drawing two face cards is Thus, P(two Kings two
face cards
3.4.11.
P(exactly one King at most one King " # % # & (' ( )
3.5.1.
!$
Figure 9.
P * # & ,+ - .+ / ,+ - (0 & .+ 21 )
3.5.3.
a) P(both are Aces 3 one is an Ace#
&
. 45
6 4 7 8 9 6 : 46
;
8
5 5 <
F KL M L
F KN F M N F K F
F O G
QR
P
@
9?
?
K S M S K
Q
QR Q
T K S
P
Q PR Q
P
4 A 9
4
@ B
@
@ B
9? C D ? 4
4
@ B ? @
@ B
?
?
?
?
?
? @
7
8E F G
U V
d) P(one is AS H both are AcesI J
J
J
3.5.5.
P
Equation 3.5.22 becomes P W X Y Z [ P W X Y \ ] Z ^ _ ` P W X Y a b c d e for f g h g i j where
e k
l m n and o p denotes the event that the gambler with initial capital q is ruined. First,
14
equation P B C D
conditions P O P
Q R
Consequently,
D
F = >
?
S
H =
and P O P
U R
I KJ L
D
;
I KJ L M
and
I KJ L
V N
Hence,
and
Z
Z [ \] ^ _ ` a
d ef g h ij
starts with dollars and stops if he reaches dollars. If that is, the game is favorable
for our gambler, then and so the gambler may play forever without
k
d ef g h ij
getting ruined and the probability that he does not get ruined is
3.5.7.
P P PP PP P
3.5.9.
P P P P PP PP P P P
blue, and
The
hit-and-run
taxi
was
black.
Then
P
Thus, the evidence against the blue taxi is very weak.
15
4.1.1.
The p.f. of
is given by
for
with histogram
0.4
0.3
0.2
0.1
!
"
if
if
if
if
if
if
if
# # #
% & &
' (
' ) '
' ' '
$
$
# $
"
"
"
#
&
" *
& $
*
"
$
+
with graph
1
0.8
0.6
0.4
0.2
4.1.3.
The possible values of are
, # ,
and -
16
.
P(2 heads
/ 01 2 3 14 2 0 5
heads
and
P(0 heads
P(4
0.4
0.3
0.2
0.1
-4
-2
5
if
if
if
if
if
if
with graph
1
0.8
0.6
y
0.4
0.2
-4
4.1.5.
The possible values of
P
)
%
"
-2
are
and
P(1 or 4 heads
$ % '
(
17
"
$ &(
' %
%
( *
P(2 or 3 heads
and
P
"
$ %& '
$ &(
' %
P(0
or 5 heads
$ &(
' %
The histogram is
(
( *
0.6
0.4
0.2
if
if
if
if
with graph
1
0.8
0.6
y
0.4
0.2
18
4.1.7.
"
) *
# * ,
&
56 7
98
for -
for 6
Thus, in general,
/ 0 1 0 2 0 3 3 3 3
/ 0 1 0 3 3 3 0 : 0
with histogram
0.1
x 4
;
56 7
?
AB C
D E
.
>
F
19
if
if
if
F
B
G
I
/
B
E J
"
# ! $
with graph
1
0.8
0.6
y
0.4
0.2
4.1.11.
First, we display the possible values of
dice:
E D
D
D
E D
D
F
x 4
D
D
E
E
E
E
E
is
20
if
if
if
if
if
if
F
0.2
-1
x 4
E D
E D
F
D
D
D
if
if
if
if
if
if
if
E
E
E
E
G
I
with graph
1
0.8
0.6
y
0.4
0.2
x 4
4.1.13.
Since is a nondecreasing sequence of events,
and
the terms of the union are disjoint, Axiom 2 gives P P P ! " # $
%
%
By the de nition of in nite sums, the expression on the right is the limit of the partial sums,
21
that is, P #
P
P
Applying Axiom 2 again, we
get P
P
P
4.1.15.
be a sequence of real numbers decreasing to
and let
Let
for every
Then
P
and
for
Furthermore,
because there is no
for which the real number
can be
for every
considering that
Thus, by the result of Exercise 4.1.14,
P
P
Hence, by the theorem from real analysis
quoted in the hint,
4.1.17.
Consider any xed real number
and let
be a sequence of real numbers decreasing
to
and
for every
Then
P
and
for
Furthermore,
Thus, by the result of
P
P
By a theorem from
Exercise 4.1.14,
real analysis, if
for every sequence
decreasing to
then
that is, is continuous from the right at
Since we have proved
this result for any real number
is continuous from the right everywhere.
4.2.1.
if
1.
Let
Then
if
or
= > ?
"
&
'&
) *
(
X Y
W
Z [
J >
J D
) /
g p
0 1 2 1
w p
4 5 6
? R
T C U
F G
C O
O O
J G K
>
c _ d
{ y
2. Hence
Thus,
<
: ;
c u
y z
f g
m n o
, -
m n o
g p
m n o
if
if
or
22
0.5
0.4
0.3
y
0.2
0.1
-1
3.
if
if
if
1
0.8
0.6
y
0.4
0.2
-1
4. P
5. P !
4.2.3.
1.
Let
C
&
> DF
DE
"
'
H I
2.
Hence K
Thus, H
LM N G
P Q
U
if
if
)( *
.
P
"
!
Then
"
#
3 54
2
4
6 78 9 : 8
J I
if
if
23
$
%
3 5;
4
3 4
: 8 <
>? @ A
=
1
0.8
0.6
y
0.4
0.2
3.
R
if
if
1
0.8
0.6
y
0.4
0.2
4.
5.
P
P
!
#
"
"
# $ %
,!
,$ %
&
4
*)
# $ %
' (
( # $ -
"
10
*) +
!
# $ %
&
( # $ -
/' (
&
# $ 0 %
1 -
' (
*)
*) +
4.2.5.
1.
2.
3.
4.
5.
6.
7.
Roll a die. If the number six comes up, then also spin a needle that can point with
uniform probability density to any point on a scale from 0 to 1 and let ! be the number
the needle points to. If the die shows 1, then let ! % ' 2 and if the die shows any number
other than 1 or 6, then let ! % # +
P ! " ' 3 # $ % )* 2
)
P ! " 4 3 # $ % 5) 2
P ' 3 # " ! " # $ % 5) ( )* % 6) 2
)
P ! % ' $ % 5) ( 7) % *7) 2
P ! . ' $ % ' ( 5) % 5* 2
P ! % # $ % ' ( 5) % 5 +
24
4.2.7.
if
if
if
if
1.
&
2.
3.
4.
5.
6.
7.
P
P
P
P
P
P
$ %
4.3.1.
The p.f. of and the possible values of can be tabulated as
!" # $
%& %%
%& %%
%& %%
%& % %
%& % %
%& % %
%& %%
%& %%
%& %%
%& %%
%& % %
,-
./
%/
%-
* .
* .
%-
* .
%-
%/
./
,-
.& %%
.& % %
.& % %
.& %%
%& %%
%& %%
%& %%
4.3.3.
The possible values of 3 are 0 and 1, and so the possible values of 0 are 4 5 6 7 4 8 (
% ( 9
1 2 !' # ( = > ? @
if A B C
: ; < Thus,
4 5674 8
if A B D ? E F
>? @
if A N C
L C
Hence, G H I A J B K > ? @ if C O A N D ? E
M
if D ? E O A F
>
4.3.5.
if Q N C
L C
G P IQ J B K >
if C O Q N > and so G H I A J B P I S O A J B P I T U V W X Y Z P [ V
M
R
if > O Q
>
^ _
if X a b
\]
Differentiating both sides, we get the p.d.f. as f g
[\ ] Y Z ` c
if X d b e
if X a b
\]
`
b
if X d b
e
4.3.7.
b
if X
^
g [ X Y Z P [ h W X Y Z P [ iV i W X Y Z `
P [ j X W V W X Y Z k ml
if t
l n o pq r s q
25
and
W \] Y Z
[X Y Z
a b
u v
Thus, $
Hence, K L
. So, if
&" '
MG N
4.4.1.
The values of
P LQ
"
MG N
if
if
u v
otherwise.
k p _
l p
m p
n p
o p
q p
f
f
f
f
f
d
l
l p j
m p k
n p l
o p m
q p n
m p _
n p j
o p k
q p l
r p m
o p _
q p j
r p k
s p l
r p _
s p j
j _ p k
n p
o p
q p
r p
f
f
f
f
q p
r p
s p
u vw
xy
f
f
f
j
k
s p
j _ p
p z {
of
j _ p _
j
k
and
j j p
j j p j
j k p _
e
j
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
_
j | l o
_
j | l o
j | l o
_
j | l o
_
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j | l o
j _
j | l o
_
j | l o
_
j | l o
j | l o
j | l o
_
j | l o
j | l o
j | l o
j | l o
j k
j | l o
j | l o
j j
4.4.3.
t
j p j p j p j p j p j {
a
} ~
2.
1.
and e
g h i
P&
4.4.5.
For
26
j | l o
" #
1.
$ % &
'
)
(
. 1
B DC
B FE
G H I J G J H
. 5
2 3
M FN
O P Q
I
d
a
e f
T
R S
. 5
: ;
<
>
U V W U
<
W U
S V\
^
`
^
^_
S[ ]
Thus,
` c
2.
g h
ij k
if
l nm
m
o
pq
r s t u s
v
and
w x
| } ~ }
3.
otherwise.
If
and
4.
then
then
and
and
then
If
#
and
27
then
otherwise.
If
and
If
$ %
"
&
'
&
&
+
4.4.9.
&
'
&
&
"
"
./ 0 1
"
'
"
&
&
"
&
"
&
"
# -
'
&
"
&
"
+
"
'
'
&
&
&
'
&
'
&
. 24 3 0 . 3 6 5 0
6
. 76 0
:8 ;9 <
"
and = >
. ? 0 1
.230 .350
2 6 2
. 76 0
:
:8 ; @
"
Now,
:8 ;9
A BC
C D
E
F
GB C
B H
4.5.3.
I
I
and J are not independent: By Example 4.4.5, and J are both uniform on the interval
KL M N O
I M
and, by Example 4.5.2, P J Q is then uniform on the unit square and not on R
H
4.5.5.
1.
2.
c U c V `
By dea nition,
l
c U f g
q
hi j
m
p
s t
l
Similarly, s t h i j u s g h i j v s t g h i j k
p o q w where the three constituents
h i j v s t g h i j is l u x v x k
l for i n
o q w
q , and l u l v l k
l for i n o q r Thus, by
s t hi j u s g hi j v s t g hi j k
l w which is equivalent to
because o
hi j
for
hi j
u
i
s g
o
n
l
o q
q r
p
28
o q
s t f g
3.
s t
s g
s t g r
If and denote the arrival times of Alice and Bob, respectively, then they will meet if
and only if ! " # $ or, equivalently, ! $ # " #
%
$ & Now, '
( " ) is uniformly
/
distributed on the square * + ( , - . * + ( , - ( and the above condition is satis ed by the points of
the shaded region, whose area is 0 1 ! the area of the two triangles 2 $ , ! 3 2 4 & Thus,
P ' 5 and 6 meet ) 2 87 9 &
29
4.5.9.
Figure 10.
Let the circle have radius and choose a coordinate system with origin at the center of the
/
circle and so that the rst random point is 5 ' ( ) & Then if the second random point is within
a distance of the point 5 ( then it must lie in the intersection of the original circle and another
circle of radius centered at 5 & From elementary geometry, the angle 6
is
and so the
area of the sector
is
The area of the triangle
is
! #"
& '
( ' *)
,. - /
$
Thus, P(the two random points will be nearer to each other than
B
C
1 2
5 64 7 8 9
: ; =<
>
?
;
C
D
E
B F
4.5.11.
1.
J K L
and
2.
b q
d r f
JN
K L
Q Q R S T U
vu
w x y
{ }|
~
V W
X Y Z V [
X \ Z ] Y ] \
_ a`
b c
de f
g h ai j k
b l
d m f n m o
if
if
otherwise
otherwise
if
and, from Part 1,
otherwise
and
otherwise.
and so
30
3.
n e
if
and
4.5.13.
From De nition 4.2.3, P
P
. /
*+
P
P(all $
! ! " #
! ! " % ! &' ( )
%
, ' ! P 0 1 2 3 4 5 6 7 8 9 : ; < P(all = > ? @ A B C D < A E F F A G H C A E I I I I J K E
1.
2.
3.
4.
4.5.15.
fg h
M N O PQ R
u v w x
if
if
X Y Z [ \
i kj l m n o p q l m r o s
for
a b c b
z |{ } ~ } ~
} ~
if and, clearly, if
4.6.1.
The joint distribution of and is trinomial (with the third possibility being that we get
any
for The table below shows the values of this function:
4.6.3.
By Example 4.5.4, P
P
P
! " # $ %
31
#
if
if
'
" #
&
(
) *
+
) 0 1
* 3
P
if + ,
if /
* otherwise.
.
.
1
4.6.5.
&'
!
/ 0 1
+ , -
6 7
3 4
if 8 9 6 9
otherwise.
if ! " "
otherwise
if
otherwise
and
?@ A 6 7
=>
if G H I J K L M
Note that O P J is the length of the horizontal line segment inside
otherwise.
the triangle at the height J I and so the conditional distribution of Q given R S J turns out
N
to H S O P J I as expected. Similarly, or by
to be uniform over the interval from H S
if ^ _ ` a b c d
symmetry, we obtain T U V W X J Y H K S Z e [
otherwise.
[ \ ]
4.6.7.
D E F
N
1.
i jk
_ b
for
e
h
and
if { |
if
if {
p q rs
f i j_ b
Hence,
Pf _
i b
Pf u
By Equation 4.6.28,
if
and
if
otherwise
w
y
{
z
32
as
if
if
if or
_ b
Thus,
.
" # $%
&' (
) *
+
-
/
7
8 9 :
B
if
if
otherwise
if 0 1 2 3
if ; < = >
otherwise
; ?
Also, by symmetry,
!
6
@
>
A
C
Q
O
2.
IH J
F G
PD K
<
L J M
<
; F G
N
P
and so,
\ ] ^_
`a b
differentiation, { |
}~
Pd]
e f gh e i j
i k l
n
c
m
o
s t
s
u s t
33
if
if
if
q v w
if
otherwise
T
U
T
U
T
U
q
p
r
x
U
W X V
p
q r
s
s y
R Y
if
if
if
R
Q
R
S
Z
[
Hence, by
Q
R
X
5.1.1.
5.1.3.
Substituting
. * /
, -
we get
~ |
5.1.5.
From the hint,
e f g h d l
results in
op q k
and
d l h m
and inte-
GF
CD E @
j
6
7
8 9 : ; < = 8 >
4
' 5
H I J K L M N O QP R S U T V W X Y Z [ \ ] ^
12 3 -
! "
$ % & ' ( )
Furthermore,
Adding the two sums and their
+ ,
by rearrangement.
5.1.7.
The distribution of a discrete . is symmetric about a number / if all possible values of . are
paired so that for each 0 1 2 / there is a possible value 0 3 4 / - and vice versa, such that / *
- :
*
6 7 0 3 8 9 For such .
7. 8 5
0 1 5
0 3
/ and 6 7 0 1 8 5
;
all 1 0 1 6 7 0 1 8 5 ; < = > ? @ A B C @ A D E
F
F
F
F
is not a possible value of P .) In the
B C
D E
G H I J ? @ K B C @ K D L (Here B C
D
M
N O if
last term, we apply the symmetry conditions: G H I J ? @ K B C @ K D M G H Q > ? C R F S T U V W X T U V Y
Thus, Z X [ V \ ] ^ _ ` a b c d e f g h i c d e c h j k l m ` a c d e f g h i k l m n a c d e f g h i c d e c h j
k
all
g c d
ef g h
g d
all
ef g h
c o
5.1.9.
By Theorem 5.1.3, p e q r s t u wv v x r y z x
the solution of Exercise 5.1.3, we obtain
5.1.11.
s {
5.1.15.
For a binomial
leads to
If
r } ~ w
5.1.13.
Example 4.3.1 gives, for continuous and
So,
, where
If
then chang-
! " # just as before.
- 0
&' ( ) * + ,- . /
- 2
, 1
Hence,
# $ % &' ( ) * & + ,
+ ,./
+ , 0 1 + , + 2 + , 3 4 5 6 76 8 9 : ; < = 4 5 6 >
!
"
5.1.17.
In this case, Theorem 5.1.6 does not apply, because ? and @ are not independent. Nevertheless, Equation 5.1.52 may still be true, and we have to check it directly:
By Equations 4.4.14 and 4.4.15 and by Theorem 5.1.1, A 7 ? : 4 A 7 @ : 4 B > On the
other side, by Theorem 5.1.4, A 7 ? @ : 4 C ED C ED
O P P Q R S T R TS U
V
V
X W [ \
O P W X Y WZ X W [ \ ] ^ _ ^ _ ]
Z
{ |
v w
5.1.19.
r x y wz s
if
otherwise
t } r }
n l m |n o p q r
and
t u
F G H IF J G K L F L G
j k l mn o p
gh i
ba c da e f
w
5.2.1.
5.1.21. r
y zq r qy s
Using the hint and the formula for the sum of a geometric series, we have, for
z s
if
This function is
otherwise
Now,
If we choose
then,
does exist.
5.2.3.
First,
for
and here the two s cancel. Second, by Theorem 4.5.4, and are independent, and so, by Theorem 5.2.3, ! " # $ % &
! " # % $ ! " #
% ' Third, by Theorem 5.2.1, ! " # % & ( ! " # % and ! " #
% &
( ! " #
% ' Thus ! " # $
$ ) % & ( ! " # % $ ( ! " # % '
5.2.5.
1.
2.
&
+ # #
, - . /
6 0- , 1 . 2 3
0- , 1 . . 2 3
9 2 / : /
H F
6 07
, - .23
08 07
, - . 0- , 1 . . /
0- , 1 .2 ;
> ?@
and then L
5.2.7.
By Examples 5.1.4 and 5.2.4,
A BCD E F
MNO
NO P
I GPD E J
NV P
35
Q D J
I G J
K F
R S T NO P U
XW Y Z
[\ ] ^
[` ] ^
X] a Y
and
1.
Thus,
and
2.
3.
&'
>4
7 ? ;
4.
LM
N O
5.
u vw
z { |
&'
>5
) %
Now,
A HB
AI B J
WX
Y Z
WX
] Z
and
and so,
*
Thus,
5.3.1.
Let us write
N
R
G H
/ 0
.
1.
M N
1.
O L
M P
2 34
Y Z ` Y
e _
a Zb
f c
6 7 8
:;
34
6 <7
"
# $
q r s
i m
d e f g h i j k d
and
4
o p
1 5 .
Thus,
3 3
Thus,
6.
! " ! #
7 8
4
(
9
! # & '
:
;
<
Now,
3
Furthermore,
(Alternatively,
: = : >
?
: > A B
C
DE
T U
and
P
R
Then
YZ
a Zb
X d
K
and
pendence of
follows. Now,
\b
a cb
X
_
1 5
WX
y
Hence,
K
J
_\
z x
5.2.9.
vw
&'
0 1
AD E
and
w
and
,+- .
*
AD E
U V
and T
}
&(
A@ B
7 ?
QN R S
Y Z
\
_
YZ
a \b
e _
a Zb
a \b
f c
a \b
a Zb
\b
a Zb
]
Y\
a \b
\b
Zb
h g
5.3.3.
k
l
Ym n
q r s tu v
w x y z
q r u sv
r x s
{
v
r x s ~
| }
m n
v
5.3.5.
36
"
$ (
# $ % ! & '
) *
+ ,
( - .
0 C 2
E F G HI
L M
+ 1
0 9
) + ,
P Q R S
Q T
R U
n s n
- . 2 3
<<
7 8
) + 1
0 ? 2
>
P Q R S
+ 1
) + ,
- . 2 3
4
+ ,
- . 5
P Q T
R U
X Y Z
X [
\ ]
5.3.9.
` a ]
d a e
i j
5.3.11.
Let
}
}
y z {
and so,
x
}
and
5.4.1.
uv
u w z
l m
n o p
denote the points showing on the three dice, respectively and let
Then, by Exercise 5.3.6,
for each and
~ }
&
! "
"
#
5.4.3.
1.
Let us write
)
/
0 1
2 7
2.
8
B
< @
4 7
C D
>
'
)
7(
7(
Then,
0 1
2 3
4 5
?
7(
<
>
< @
C ?
< @
>
For instance, P
and are not independent.
G H
and
I J
K L
but P
M N
G H
E L
PO
A
O
M N
5.4.5.
1.
With
b
c d
e f
T
a g
W X
V
U
h
W [
W X
V
~ {
_ X
^
T
e
W _
_ X
^
U
[
_ ` _
m n o
or, alternatively,
m t
r s
u v
y
z {
2.
and
5.4.7.
sign
37
and
Thus,
5.4.9.
By the result of Exercise 5.4.1,
and, by Theorem 5.2.2 and by the obvious property
? @ A B CD E F GH I J K L M N
5.5.3.
Clearly, a
5.5.5.
Here,
[b
_
z
Equation 5.5.10,
if
if
+2 *
W X
Thus, a
5
Z
#"
if
otherwise.
S
R
S
Thus,
\ ]
* +
&
,-
g h
f q
c
t q
s
v w
)*
: =
nm
j on
qr s
W X Z `
and
we get
and, by
+ & ,- .
8; 7
8<
[b
k w
t ~
: :
^ c
otherwise. From
and so,
#"
[X ^
Thus, ;
j k l
& '
./ 0
directly as
if
if
otherwise
6
=
> ? @
8
E
<
F
and
if
otherwise.
if
if
otherwise
U
R
f
e
f
8<
[
d
hi
NF
[X \ ] ^
89 : ; 5
6 7
KL
5.5.7.
I J
Hence, for
. Now,
and
[b
. .
+3
x w v
otherwise. Similarly,
and
We can also compute
By symmetry,
too.
hg c
and
[b
O L TP Q
SN
if | } v w |
otherwise.
t uv w x y
and a
OP Q M R
and so,
5.5.1.
By De nition 5.5.1, for discrete
and
Theorem 5.1.3 with / + 0 . 1 2 * + 3 . , 2
>
and
5.5.9.
By De nition 5.5.1 and Theorem 5.1.4,
and
38
5.5.11.
From Theorem 5.5.1,
thermore,
U
V
1 2
T
U
V
1 6 7 7
8
3 4 5
P L
^:_ Y
\e
^ g
~
and so
Let
P
such
for
2.
^ _ Y
9 ;:
9 ;:
:
Y- f
^ b
c d h
\ e
& ' ( ) * +
1
:
K L M& ,N O' * +P Q- R. K/ Q
A B C
j kl
D E F GH I
EH I
p l
q r s t
On
in general.
and since
for all
Thus, the median is
any number such that
and P
and
Then P
for all
is a median.
\]
and P
for
>
\ e
1> 7
c d
Let
= 5
! "
# $ %
1.
with density
}
yz
1 6
` ] ` _
D J
w
x w
3
4 <
;:
\]
Fur-
5.5.13.
For continuous
v
X Y Z
Now,
Then
Thus, any
5.6.3.
P
The converse of Theorem 5.6.1 says: For a median of a random variable
and P
imply P
This statement is true, because one of the
statements
and
is certain to be true and so P
P
P
On the other hand, the hypothesis says that P
P
and subtracting the latter equation from the former, we get
P
5.6.5.
hence,
using the Fundamental Theorem of Calculus, we get
Thus,
has a critical point where
or where
that is, if is a median . Since we assumed that is continuous and
is unique. The second derivative test shows that
has a minimum at
, for
by assumption.
5.6.7.
For general
the 50th percentile is de ned as the number
Since any d.f. is continuous from the right, P
and so we have
" $#
% &
'
( )
+ ,
- ,
>
'
+ 23
2)
0
5 6 7
9 :
< :
/ 0
A B
D B
A G C H
IJ
A G C K
A NO
P F
NC
A G C H
A B
A G C
T
S
A NO
NC
W?X
A NO
NC
P @
A G C
f l
39
no
\]
f c
no
` a
b c
ij
f c
ij
P
such that
f l
no g
"# $ %
&' &
no g
5.6.9.
0
1
ij
M N J O
,- . /
for
IQ R
i j
if - 4
if > ?
if C D
C
Its graph is
78 9 : ; <
=
2
C
5 6
@
A B
A E
-1
5.6.11.
From Equation 4.2.15,
F G H IJ K L
T
S
U
VW
if
if
if
]
VW _ `
40
W Z [
\
X Y
[
W Z ^
\ Y
\
^
W
]
\ Y
Y
41
6.1.1.
1.
]
is Poisson with
] _
! "
2.
3.
P7 8
4.
is Poisson with #
9 7:; <
and 8
PO P
Thus P
= 7:; <
]
] _
P
]
! " % " $
?P 7 8
7:; <
:;@
)/ ., -
A: B 6 C D EFG H
U WP
]
) 0, 1 2
)
I JK L M N J
OU S T X S T
V ORS T
P
]
Thus P !
:; >
]
P Y Z [ Y Q \ ] ^ and _ ` a b \ ] ^ \
d
Pa_ a ^ \ ] c \
and P
Q ORS T U
p qr s t q
]
3 45 6 5 4
l
e fm g h i h j j
jk
g j ijn
nk
6.1.3.
u v
.per x
y z.
w
1.
P{ |
{}~
2.
3.
4.
5.
P
P
6.1.5.
P(even P(odd
3
P
P
!
"
$
%
&
'
( ( ( )
+ , - . + , - .
*
+ , / 0 1 2
6 7
jk vl
j o v m
jo vl
t o vl
3 5
p
u l
q r n
jk v s
xy z { | } ~ } ~ | } ~ } ~
t k vl
jk vl
u wn
jo l
9 :
P(odd
and so, adding the two equations, we get 8 P(even
3 5 6 ; < = > ? @ A
ing them, 8 P(odd
6.1.7.
; C D
E
D
E
F
E
F G C F
H
D I J
C D I E
D I
Consider the instants B
and let K L and K M denote two distinct interarrival times. Then N O P Q O R
] ^ _ ` a b c d e f b c d e a g b c d e f g b c e f he f g
P(i j k l m n j o l p q r n j k s t k l m n
n
jo
/ 0 1
ST
F I
jk l
t o l
, and subtract-
F I G
V W T X Y
p
C F I
Z X[
j o vl
u r
u r n
| |
1.
42
2.
3.
4.
5.
6.
7.
8.
9.
10.
P
P
P P
P P
P
P P
P P P
P
P
P
P
P
P
6.2.3.
1.
B D E F G
changes
sign at
B E H FI
2.
that is,
23 4 5 6 7 8 9 : ; < = > ? @@ AB C
! " # $ % & ' ( ) * * + , - . 0 /
B E H F I
'1
K
at
Thus, J has points of in ection at and only
at
k o p g k q r ss t u v w
PR
N
O P Q R S T UV T W XY Z
Y [ \ ] ^ _ ` a b d e c f g h gi k j l i m h i j n
}
x z { y| } ~
changes sign at that is, at Thus, has points of
6.2.5.
1.
; <
2.
T S
<
E
D
I
F G
V \ ] ^
! "
#
$ ! % & $
'
) % *
+" + &
, - ./ - .0 1 2 3 4 4 5 6 7 . 0 8 4 5 9
>
;
K [ \
I Z
I J
K L
? ; @
MN O
` ab c d
e f g d
A B
I J
K L V W
R S T U
B CL
T Q
R S
T X
M Y N O
ab h f i g i ] ^ j
k l
^ j
o k p
6.2.7.
Comparing
with the general normal p.d.f.
and
this distribution is normal with
6.2.9.
and
denote the two weights. Then, according to Theorem 6.2.6,
Let
with
and
Thus, P
P
P
` s tu v w x y z w {
s tu s x y z w y
} w|~
43
is normal
6.2.11.
then, since is strictly increasing, we can solve this equation for to
If
get
or
Here
is the area of the tail to the right of
under the standard normal curve, which equals the area of the corresponding left tail, that is,
So,
Solving this equation results in
, which,
when we substitute from the rst equation, yields
6.3.1.
and
:P
We use the binomial p.f. with
0 ' &1
( &1 5
C D E FG L
CE F K L D
OK
' %
6 78 9 : ; < =
D
, !
>
C E FG L P
8 < =
CE F K L Q
!
' %
C D E FG
"!
"!
E F K L
% &' ( ) &
"
and so,
C E F K L
C E FG L D
and
= 78 9 : ; < =
8 < =
N
E FU S K V D
E FG R S T Q
E FW R K W F
B
6.3.3.
and
By CorolA single random number is a uniform random variable with
lary 6.3.2, the average of
i.i.d. copies of is approximately normal with
[
^ [ g g
\ e
1.
We want
2.
^ [ _
b c c
d e
and
P
6.3.5.
[Z
Thus, P
i ji k l j
mi jn l
table, amounts to
or
i jq r s
u v wx y z { w|
z { w|
{ w{ } y
{ w | z { w|
{ w { } y
{ w{ } y
and
Equivalently,
and so
and so we want P
6.4.1.
successes will occur before failures if and only if the th success occurs at the th trial,
for any
. Thus, using the negative binomial p.f., we obtain P( successes before
failures
6.4.3.
If the number of failures before the th success is then the total number of trials up to and
including the th success is
Thus, P
P
where
is negative
binomial, and so P
6.4.5.
P
P
P
for
and
6.4.7.
# $
@ >
? A
< B >
C >
+ '
>
( ,
. /
& '
& '
1 2
5 3
A D
E F
G H I J
K E F
G L
! "
6 7
& '
< = >
P O
R P O
44
6 7
1 9 3: 2 3
2 3
P M
: 8
9 3
R P T
Letting
8 9: = > ;
+ 0 1 2 0
1.
! "
N O
PQ
) '* (+ , - . /
' 3 4 -
lm n
8 69 :
7 ;
Since
: 9: = A B CCCDE F G H A B
<
I J
L
S
UQ
V W
YT
o qp
R ZV
PQ
v y z { y u
3.
6 7 ? @
integer K
2.
$ &%
t ru v
s w x
a ^ _
[ \ [` ]
c ` d e
[
hg i j
y |w } ~
for
6.4.11.
for
We prove
it reduces to
by induction. For
Equation 6.4.30, which was proved in the book. Next, assume that it is true for
Then, using also the reduction Formula 6.4.11, we get
Thus, the
truth of the formula for any implies its truth for and so it is proved
for
any
6.4.13.
and are i.i.d. standard normal variables, and so is with 2 degrees of free
dom. Thus, P P
Hence,
if
, which shows
if
that is exponential with parameter
In particular, the
distribution is the same
as the exponential with parameter
6.4.15.
if
if
and
By Theorem 4.5.8,
if
if
if
Thus,
if
if
if
and
otherwise
otherwise
Comparing these expressions with De nition 6.4.4, we see that is beta with
and
and is beta with
and
.
6.4.17.
8 9 :
7
# $ %
"
&
'
)( *
*
- * .
,
1 02
4 5 2
2
8 <9
>
?
F G
C D E
Q RP S
E I E
H
QP T
XY Z
[V \
X Y Z ]^
c e
b
g
pg
i q
k l m r e
k g
y ~
d
l
b
g
l m e
b
g
d
c
b
g
t u
kc m
w
v
xy z
45
k l m
1.5
0.5
0.2
0.4
0.6
0.8
1.5
0.5
0.2
0.4
0.6
0.8
46
0.5
0.4
0.3
0.2
0.1
0.2
0.4
0.6
0.8
1
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
7
6
5
4
3
2
1
0
0.2
0.4
0.6
0.8
47
4
3
2
1
0.2
0.4
0.6
0.8
6.4.19.
In Theorem 4.6.2, Equation
4.6.26, we )substitute
* + , - . / 0
for
# (
if 1 2 3 2
otherwise
and
1
G H
J K L M N O P M Q
ST U V W
T U V U Z Z Z U [
U
for
and X Y
where we
otherwise
]
left the constant \ undetermined. Its value could be determined by the coef cients in ^ _ ` a
and b a and the integral in the denominator of Bayes c Theorem, but we can d nd it much more
easily by noting that the variable part being a power of e times a power of f g h i the posterior
density j k l m must be beta. Thus, j k l m is beta with parameters n o p and q g n o r i and
6 7
89
:; < = >
u
v w x y z {| } x y ~
6.5.1.
Clearly, u and as linear combinations of normals, are normal. To show that they are
standard normal, we compute their expectations and variances: u
and
because
Now,
because
)
) * + ,
@
Z
1, 0
- . /
,
10 0
0
+
3 4 5
9
6 7 88
8
and
! "
#$ % & '
9
7 ;8 <
:
; = > ?
A
R W X
R Y
A C D E F G HI G H
H JKL M N O M P Q R S T T U V
Q T S R R P Q R S T R U V
e fg e h i j
i
i
h i
i
h
[
a
T
j j
j j k l m n o n j p q j o
o o
o j r
T \ ] ^ _` ^ ` ^
^ bc d
d
d
B
48
(
j
j
n j p
si j
h
o o
i j
oj k t
Also,
o
Similarly,
$
! " #
! %&
> ?@ A
(
2
2
<
B ? C D EF G G
/ 114
(* +, -. /01
3
1 5 6 7 standard
8 9 : ; ; normal,
and so, using the fact that H I and H E are0 independent
b
e f
; =
J K L MN
N
T
I
U
R
I O E P Q
Y
Y
Y
a
a
a
I S T U VW XX
W UX [ W X X
^ _`d
X Z
X \ ] ^ _`
^ ``
g i
h
_
` c
_
r
r
r
h j k l m ln m l
l op q p r s t q q t r q u t q r t r r v w p r x t q q u t q r y z {
| }
q
6.5.3.q ~
' )
Equating the coef cients of like powers in the exponents in Equation 6.5.14 and in the
present problem,
we get, for the coef cient of for the coef cient of
and for the coef cient of
These three equa
and
constant
Furthermore,
Now,
and
! "
"
parameters.
6.5.5.
By the result of Exercise 5.4.8,
&
'
#
! )
, . /
! "
% 3
4 5
7 8
7 ; <
>
A B
:
8
; <
A B
Hence,
By Theorems 5.4.2, 6.5.1 and 6.5.4,
and
are independent if and only if their covariance is
zero, that is,
This equation is
equivalent to
6.5.7.
The conditional expected score on the second exam is given by Equation 6.5.9 as
The conditional variance of
is given by
Equation 6.5.10 as
From the table, the 90th
percentile of the standard normal distribution is
Since
under the condition
is normal with
and
we obtain
J
>
A K
; <
A F
A B
A B
E E
>
A B
; <
A F
g h
jh
A B
Y Z
V8 R
OS P
T UQ
[ \
] E^ _
[ \
b c
Y Z
A B
[ \
d\
` ^
E E
g h
w x v
k i
>
A B
jh
| }
49
6.5.9.
is bivariate normal as given by De nition 6.5.1, then
is a linear
If
combination of the independent normals
and
plus a constant, and so Theorems 6.2.4
and 6.2.6 show that it is normal.
To prove the converse, assume that all linear combinations of
and
are normal, and
choose two linear combinations,
and
such that
Such a choice is always possible, since if
then
and
will do, and otherwise the rotation from Exercise 6.5.5 achieves it. Next,
we proceed much as in the proof of Theorem 6.5.1: Let denote the bivariate moment
generating function of
that is, let
Now,
is normal, because it is a linear combination of
and
Denoting the parameters
of
and
by
and
respectively, we have
and
(There is no term here with
because we have chosen
and
so that
Denote the mgf. of
by
that is, let
Then,
$ %
&
'
"
&
&
( )
&
&
- .
- .
$ ,
'
&
& 0) -
&
4 (
'
&
2 3
& - %
) 5
+5
by Equation 6.2.15,
4 >
E H EI
: ; <
E H E
F O
M P
G D
$ -
E H EI
K L
C D E B
E K L
b c
v wx
; ? @ A B
C D E B
E H E
F O J
h e
v u
x z
Y ^ ]
k l
m n
q r
and
h e
^ X
Y ^ ]
k l
Thus
Z `
m n
50
x
y
_ ]
E J K L
x
u
F G
7.1.1.
Replacing
!" # $ %
&
' (
) * +
'
,
-
$
<
< =
>? @
X
AB C D E
\
]
IH =
cd ` e
_` a b
I =K L
f g h
'
12 3 4 5 6 2 7
NO P Q R S O T
which is kj
.
- /
U V W
Y Z
Hence,
8 9 :
h l
7.1.3.
If m n is a discrete r.v. with o possible values p q and p.f. r s p q t u wv for all x y then, by
equation 5.1.5, z u { s | n t u } ~
and, by De nition 5.2.1 and Theorem 5.1.3,
7.1.5.
a)
and so
7.1.7.
By Theorem 4.5.8, if
d.f.
then
%
&
'
( )
+%
&
( )
* ,
?
<
@
=
>
"
!
C D
s t u
'
( )
/
.
{|
~
|
{|
S
P
T
Q
X [ Y\
` a
b
which
i
h
b
j k
b
l
z z
and
We are given the successive values
to obtain
the equation
and
vals are
g n
and
Therefore,
shows that is an unbiased estimator of
7.1.9.
This problem is an instance of the general case considered in Example 7.1.8. Here
8 4
and so
#
Hence,
at
Hence,
yields the critical value
,
and
, for each of which we need to solve
Hence,
Now,
So, the required approximate con dence inter-
and
7.2.1.
We use a large-sample Z-test. The null hypothesis is that the sample was selected from the
student population with mean grade 66 and SD 24, that is,
is
The alternative
is that the students in the sample come from a different population, for which
The test statistic is
which we take to be approximately normal, because is suf ciently
51
large for the CLT to apply. The rejection region is the set
We compute the P-value
as P
P
This probability is high
enough for us to accept the null hypothesis, that is, that the low average of this class is due to
chance, these students may well come from a population with mean grade 66.
7.2.3.
We use a large-sample paired Z-test for the mean increase
of the weights, with
denoting the hypothetical mean weight of the cow population before the diet and
that
after the diet. We take
and
The test statistic is
the mean weight
We assume that is
increase of the cows in the sample. The rejection region is
approximately normal with SD
We compute the P-value as P
! " #
&
&
" % !
'
&
&
&
&
'
&
1 2
( +
6 7 8
4
: ; < = ;
> ?
A B
CD
E F
P
Thus, we reject the null hypothesis: the diet is
very likely to be effective however, the improvement is slight and the decision might hinge
on other factors, like the price and availability of the new diet.
7.3.1.
a) Here a Type 2 error means that we erroneously reject an effective drug.
b) Accepting the drug as effective means the same as rejecting
Thus, we want
P
which, from Equation 7.3.4, approximately equals
If
then the drug has really reduced the duration of the cold from 7 to 6.5
days, and the test will correctly show with probability
that the drug works.
7.3.3.
We use a large-sample Z-test. The null hypothesis is that the sample was selected from the
is
The alternative
student population with mean grade 66 and SD 24, that is,
is that the students in the sample come from a different population, for which
The
test statistic is
which we take to be approximately normal, because is suf ciently large
for the CLT to apply. The rejection region is the set
We compute the P-value as
P
P
Thus, solving this equation
N OK
H IL M J
L QM NK O
W X
YZ X [
Y] ] ] ^
e f
ij
mn o
e r s tw u v
YZ [
| } ~
t su
s x u
W c
for
is
yields
52
1
0.8
0.6
y
0.4
0.2
20
40
60
80
100
7.3.5.
Let
denote the number of nondefective chips. The rejection region is the set of integers
The operating characteristic function is
P
and its plot is given below.
1
0.8
0.6
y
0.4
0.2
0.6
0.7
0.8
0.9
This is not a very good test: For instance, when the probability of a chip s being nondefective is .8, the graph shows that the test still accepts the lot with the fairly high probability of
about .3. We could improve the test by sampling more chips or by rejecting the lot if even
one defective is found in the sample.
7.4.1.
For the given data, we nd and ! " # $ $ We use the % -distribution with 4 degrees
of freedom. We want to nd % such that P & ' ( % ) $ * + $ From a % -table we obtain % ! # $ * , and so P . / # $ * , ( 430 5 16 27 8 9 : ; < = > : ? @ : Substituting the observed values A > ? B ? and
CD E
9 :B
for G and
HC I
we get P J ? B
Q
? K 9 : ; < L M NQ O P R S R T U T V W X Y Z [ \ Q O P ] ^
53
_` a b
that is,
To nd a 95% con dence interval for b we can proceed much as in Example 7.4.1. By
Theorem 7.4.2, has a chi-square distribution with degrees of freedom. We obtain, from
a table or by computer, the and the quantiles of the chi-square distribution with
!
!
" " "
and P
Thus,
degrees of freedom, that is, P
P#
' (
% &
) (
a _
+ + , + - .
D D E D F
0 ,1 2 ,
EF D EG
or, equivalently,
7.4.3.
M
D P L EG
I
R
K D ES E
and Q
We use the T -distribution with 4 degrees
For the given data, we nd N O
of freedom. We reject U V if W X Y Z [ or, equivalently, if \ X ] ^ where ] _ de f` ga hb c
k l m no p k q m
a i
j
r k ns t u v
w
x y z z { y Thus, from a table or by computer, P | }
~
yz z {
y and so we
accept the null hypothesis, the truth of the store s claim.
7.4.5.
We can write | where is the normalization constant given ex
, we
Thus, writing
and
obtain
provided
exists. Assuming this result for the moment, we get
Now,
"
# $ &%
'( )
, & - . / 0 12 345 /
*+
` a
=
A =
89: ; < =
bc d e f g
BC D
QR S
; EAF G H I J K L M N O I
89: ; < =
>; 7
U T VW X Y Z [ \ ] ^ _
hij k l m
s t uv w x y z { |
~
7.4.7.
For
and so,
Also,
7.5.1.
We are testing
against
to be
ber of yellow seeds turned out
54
"
$
7.5.3.
as in Equation 7.5.20. For each subpopulation, the sum
The number of terms is again
of the entries is prescribed, and so the degrees of freedom are reduced by Also, the sums
in each category are estimated from the data, but only
of them are are really estimated,
because the sum of the column sums must equal the sum of the row sums. Thus, the nal
number for the degrees of freedom is
7.5.5.
We divide the interval into four equal parts (in order to have the expected numbers equal not
less than ve) and the list gives the following observed frequencies for them:
%
&
' (
'
' (
'
(
&
'
(
Intervals
Frequ.
#
# # * #
#
+ ,
+
#
+
# * #
23 4 5 6 7
0 1
/
* #
+
4 5 6 7
- + ,
#
- +
5 4 5 6 7
*
# #
4 5 6 7
Thus,
We have 3 degrees of freedom, and so
a table gives P
Hence, the calculator seems to generate random numbers
very well.
7.5.7.
We can extend the table to include the marginal frequencies:
Sex Grade
M
F
Either sex
L
9:
C
@A B
E
A B F
9;
<
H IJ K I
P P
P P
17
13
15
14
= >? >
16
Any grade
31
57
88
13
Hence, the expected frequencies under the assumption of independence can be obtained by
multiplying each row frequency with each column frequency and dividing by 88. So, the
expected numbers are
Sex Grade
M
F
L
K IJ R
M IJ J
K IM Q
M IS Q
M IN K
K IM Q
J IH O
P P IH P
Q IK S
J IO S
P H IR N
Q IK S
55
Thus,
M
E
A B
K IJ R B
M IJ J B
K IM Q B
M IS Q B
M IN K B
G
K IJ R
M
K IM Q
P P
M IJ J
K IM Q B
P P
J IH O
J IO S B
K IM Q
J IH O B
P P IH P
P H IR N B
M IS Q
P P IH P B
P H IR N
Q IK S
Q IK S B
G
J IO S
M IN K
Q IK S B
P I P N
Q IK S
P
P
@A B
E
A B F
H IJ M
0 1 23 4
9 : ; ?
! !
"
# $ !% &! '
0 5 21
) *+ ,
0 26 1
< =
8 9 : ;
>
@ 9@ @ 8 A
< C
EF
EF
J K LM
K LO
S T
V
U
m
n
Z Y[
\
o
p
q r st
u v
x v
}
~
` a
Hence,
7.7.1.
Use the large sample formula
c
b
i hj
k
if
From
we get
and so
Thus,
and, this value being fairly large, we accept
7.7.3.
Since
has only a nite number of values, it does assume its supremum at
some values of that is, its supremum is its maximum. Also, since and are
right-continuous step functions with jumps at the , is assumed at
every point of an interval and, in particular, at !
7.7.5.
$ ) %
Use the large sample formula P " # $ % & ' ( $ % * + , - /. 0 1 2 3 4 5 6 7 8 9 : ; < = > = ? From @ A B C
B
F
we get G
B D
` a
b c d
H I I C J I I
and so S
b i dH Ij I kK Ll Mm M n o pNq r sO n P O tQ R u v w x v
e f
g h
Thus, P W X
T
O PQ U V P
We accept y
56
v
z
Y Z
\ Y Z ]
D
D
http://www.springer.com/978-0-8176-4497-0