(ENM 503)
Michael A. Carchidi
December 7, 2014
Chapter 8 - Properties of Expectation
The following notes are based on the textbook entitled: A First Course in
Probability by Sheldon Ross (9th edition) and these notes can be viewed at
https://canvas.upenn.edu/
after you log in using your PennKey user name and Password.
1. Introduction
In this chapter, we develop and exploit additional properties of expected values. To begin, recall that the expected value of a discrete random variable X is
defined by
E(X) =
+
X
x=
xp(x)
or
E(X) =
xp(x),
(1a)
x|p(x)>0
where the second sum need only be over those values of X = x for which its pmf
p(x) is non-zero, and the expected value of a continuous random variable X is
defined by
Z +
Z
E(X) =
xf (x)dx
or
E(X) =
xf (x)dx
(1b)
x|f (x)>0
where the second integral need only be over only those values of X = x for which
its pdf f (x) is non-zero.
then
a E(X) b.
xp(x)
x|p(x)>0
or
a
x|p(x)>0
But since
x|p(x)>0
x|p(x)>0
p(x) E(X) b
p(x) = 1
p(x).
x|p(x)>0
we now have
x|p(x)>0
a E(X) b
fY (x)dxdy
(2)
where the region of integration in the xy plane is below the 45 line in the first
quadrant, as shown in the following figure.
The region
R = {(x, y)|0 y < , y x < }
is below the 45 line in the first quadrant
This region described by
R = {(x, y)|0 x < , 0 y x}
is also below the 45 line in the first quadrant of the xy plane, and so interchanging
the order of integration in the above integral, we have
Z Z x
Z Z x
Z
Z
P (Y > y)dy =
fY (x)dydx =
dy fY (x)dx =
xfY (x)dx = E(Y )
0
P (Y > y)dy
P (Y < y)dy =
Z y
fY (x)dxdy,
xfY (x)dx
where the region of integration is below the 45 line in the second quadrant of
the xy plane, as shown in the following figure.
The region
R = {(x, y)|0 y < , < x y}
is below the 45 line in the second quadrant
This region described by
R = {(x, y)| < x 0, 0 y x}
is also below the 45 line in the second quadrant of the xy plane, and so interchanging the order of integration in the above integral, we see that
Z
Z 0 Z x
Z 0 Z x
Z 0
P (Y < y)dy =
fY (x)dydx =
dy fY (x)dx =
(x)fY (x)dx
0
and so
Z
0
P (Y > y)dy
which leads to
Z
P (Y < y)dy =
xfY (x)dx +
xfY (x)dx =
xfY (x)dx
(x)fY (x)dx
P (Y < y)dy
(3)
or
E(g(X)) =
x|g(x)>0
f (x)dydx =
x|g(x)>y
x|g(x)>0
g(x)
dy f (x)dx =
and so
E(g(X)) =
g(x)
f (x)dydx
g(x)f (x)dx
x|g(x)>0
g(x)f (x)dx
(4a)
(4b)
and
E(g(X)) =
g(x)f (x)dx
(4c)
(5b)
To prove this, suppose that g(X, Y ) 0, then from Equation (2), we have
Z Z Z
Z
P (g(X, Y ) > t)dt =
f (x, y)dydxdt.
E(g(X, Y )) =
0
(x,y)|g(x,y)>t
or
E(g(X, Y )) =
Z Z
x
g(x,y)
dt f (x, y)dydx
1
1
1
= 2
L L L
for 0 < x < L and 0 < y < L, and zero otherwise, we have
Z LZ L
Z LZ L
1
1
E(|X Y |) =
|x y| 2 dydx = 2
|x y|dydx.
L
L 0 0
0
0
Now since
|x y| =
we have
Z
x y, for x y > 0
y x, for x y < 0
|x y|dy =
(x y)dy +
1
(y x)dy = x2 + L2 xL
2
and so
1
E(|X Y |) = 2
L
or E(|X Y |) = L/3.
1 2
2
x + L xL dx,
2
or
E(X + Y ) =
+ Z +
xf (x, y)dydx +
yf (x, y)dydx
or
E(X + Y ) = E(X) + E(Y )
(6)
for any two random variables X and Y and constants and . More generally,
we may use mathematical induction to show that
E(1 X1 + 2 X2 + + n Xn ) = 1 E(X1 ) + 2 E(X2 ) + + n E(Xn )
(7)
Xi
(8)
X=
n i=1
Xi =
E(Xi ) =
= .
E(X) = E
n i=1
n i=1
n i=1
This says that the expected value of the sample mean statistic is , the mean of
the distribution. When the distribution mean is unknown, the sample mean is
often used in statistics to estimate it.
Booles Inequality
Let A1 , A2 , ..., An denote n events and define the indicator variables X1 , X2 ,
..., Xn by
Ai occurs
1, if
Xi =
X=
n
X
Xi
i=1
1, if X 1
Y =
0, if X = 0
so that Y is equal to 1 if at least one of the events Ai occurs and Y is zero if none
of the events Ai occur. Now it should be immediately obvious that X Y , so
8
n
X
i=1
n
n
X
X
E(Xi ) =
((1)P (Ai ) + (0)P (Ai )) =
P (Ai )
i=1
i=1
and
E(Y ) = 1P (X 1) + 0P (X = 0) = P (X 1)
and
P (X 1) = P (at least of the Ai occur) = P (A1 A2 An )
we see that
P (A1 A2 An )
n
X
P (Ai )
(9)
i=1
if xi = 1
p,
p(xi ) =
1 p, if xi = 0
for i = 1, 2, 3, ..., n, so that E(Xi ) = (1)(p) + (0)(1 p) = p. Since
X = X1 + X2 + X3 + + Xn
we see that
E(X) = E(X1 ) + E(X2 ) + E(X3 ) + + E(Xn ) = p + p + p + + p
9
or E(X) = np.
Example #4: The Mean of The Pascal Random Variable
Recall that the Pascal random variable X with parameters p and r has pmf
x1 r
P (X = x) =
p (1 p)xr
r1
for x = r, r + 1, r + 2, ..., and recall that such a random variable represents the
number of independent and identical Bernoulli trials, each with a success probability of p, needed to achieve the rth success. If Xi is geometric with parameter
p, then this random variable represents the number of independent and identical
Bernoulli trials, each with a success probability of p, needed to achieve the first
success so that
p(xi ) = p(1 p)xi 1
for xi = 1, 2, 3, ..., and i = 1, 2, 3, ..., r, and E(Xi ) = 1/p. If Xi (for i = 1, 2, 3, ..., r)
are all independent and identical geometric random variables, each with parameter
p, then
X = X1 + X2 + X3 + + Xr
is Pascal with parameters p and r and then we see that
E(X) = E(X1 ) + E(X2 ) + E(X3 ) + + E(Xr ) =
1 1 1
1
+ + + +
p p p
p
or E(X) = r/p.
Example #5: The Mean of The Hypergeometric Random Variable
Recall that the hypergeometric random variable X with parameters g, b and
n has pmf
g b
P (X = x) =
nx
g+b
and
10
If n balls are randomly selected from an urn containing N balls of which g are
white, find the expected number of white balls selected. To solve this, let X
denote the number of white balls selected, and let us represent X as
X = X1 + X2 + X3 + + Xn
where
Xi =
0,
otherwise
for i = 1, 2, 3, ..., n. Then, since the ith ball selected is equally likely to be any of
the N balls, it follows that
g
E(Xi ) = .
N
Hence
E(X) = E(X1 ) + E(X2 ) + E(X3 ) + + E(Xn ) =
g
g
g
g
+
+
++
N N N
N
or E(X) = ng/N.
Example #6: A Coupon-Collecting Problem
Suppose that there are N types of coupons, and each time one obtains a
coupon, it is equally likely to be any one of the N types. Determine the expected
number of coupons one needs to amass before obtaining a complete set of at
least one of each type of coupon. To solve this, we let X denote the number
of coupons collected before a complete set is attained and let us define Xi (for
i = 0, 1, 2, 3, ..., N 1) to be the number of additional coupons that need to be
obtained after i distinct types have been collected in order to obtain another
distinct type, and we note that
X = X0 + X1 + X2 + + XN1 .
When i distinct types have already been collected, a new coupon obtained will be
a distinct type with probability 1 i/N. Therefore
k1
i
i
P (Xi = k) = 1
N
N
11
1
N
=
p
N i
and so
E(X) = E(X0 ) + E(X1 ) + E(X2 ) + + E(XN1 )
leads to
E(X) =
which reduces to
N
N
N
N
+
+
+ +
N N 1 N 2
N (N 1)
E(X) = N
N
X
1
i=1
= NHN
1, if X i
Xi =
,
0, if X < i
then
Xi =
i=1
and then
E(X) = E
X
X
Xi +
i=1
X
i=1
Xi
i=X+1
X
i=1
X
X
Xi =
(1) +
(0) = X
i=1
X
E(Xi ) =
((1)P (Xi i) + (0)P (Xi < i))
which reduces to
E(X) =
i=1
X
i=1
i=X+1
12
P (Xi i)
(10)
(11)
Z + Z +
g(x)h(y)fX (x)fY (y)dydx
=
Z +
Z +
g(x)fX (x)dx
h(y)fY (y)dy
=
= E(g(X))E(h(Y )).
(12a)
(12b)
(13a)
Note that
and
Cov(X, Y ) = E(XY ) E(X)E(Y ) = E(Y X) E(Y )E(X) = Cov(Y, X) (13b)
13
and
Cov(X, Y ) = E(XY ) E(X)E(Y ) = E(XY ) E(X)E(Y )
or
Cov(X, Y ) = Cov(X, Y ).
(13c)
j=1
i=1
= E
j=1
n m
XX
Xi Yj
i=1 j=1
=
=
m
n X
X
i=1 j=1
m
n X
X
i=1 j=1
and hence
Cov
n
X
Xi ,
i=1
m
X
E(Xi Yj )
i=1
n
X
i=1
m
n X
X
j=1
! m
!
X
E(Xi )
E(Yj )
j=1
E(Xi )E(Yj )
i=1 j=1
Yj
j=1
m
n X
X
Cov(Xi , Yj ).
(13d)
i=1 j=1
0, if X 6= 0
1, if X = 0
14
1
3
or simply
V
i=1
n
X
Xi
i=1
n
n X
X
j=1
Cov(Xi , Xj ).
(14a)
i=1 j=1
i=1 j=i
n
X
Cov(Xi , Xi ) +
i=1
or
V
n
X
i=1
Xi
n
X
i=1 j6=i
n
n
XX
Cov(Xi , Xj )
i=1 j6=i
V (Xi ) +
i=1
n
n X
X
Cov(Xi , Xj ).
(14b)
i=1 j6=i
i=1
(14c)
i=1 j>i
15
(15a)
for i = 1, 2, 3, ..., n, are called devibe the sample mean. The quantities Xi X
ations, as they equal the dierences between the individual data and the sample
mean. The random variable
1 X
2
S =
(Xi X)
n 1 i=1
n
(15b)
V (Xi ) = 2
= 2 n 2
V (X) = 2
n i=1
n i=1
n
or
1 2
.
n
To do the second part of the problem, we first note that
=
V (X)
(n 1)S
n
n
X
X
2
2
=
(Xi X) =
(Xi + X)
i=1
n
X
i=1
i=1
+ ( X)
2)
((Xi )2 + 2(Xi )( X)
n
n
X
X
2
)2
=
(Xi ) 2(X )
(Xi ) + n(X
i=1
i=1
16
(15c)
n
X
)(nX
n) + n(X
)2
=
(Xi )2 2(X
i=1
n
X
)2 + n(X
)2
=
(Xi )2 2n(X
i=1
or
n
X
)2 .
(Xi )2 n(X
(n 1)S =
2
(15d)
i=1
Then
!
n
X
)2
(Xi )2 n(X
E((n 1)S 2 ) = E
i=1
or
n
X
(n 1)E(S 2 ) =
i=1
n
X
i=1
n
X
i=1
)2
E(Xi )2 nE(X
E(Xi )2 nV (X)
V (Xi ) nV (X)
(n 1)E(S ) =
n
X
i=1
1
2 n 2 = (n 1) 2
n
(15e)
(15f)
1
E(T ) = 1
2
n
(15g)
T2 =
we find that
17
(16a)
(16b)
and when Cov(X, Y ) ' 1, we says that the random variables are positively correlated. Also
Cov(Y, X)
Cov(X, Y )
=p
= (Y, X)
(X, Y ) = p
V (X)V (Y )
V (Y )V (X)
and
(16c)
Cov(X, Y )
Cov(X, Y )
Cov(X, Y )
p
=p
=
(X, Y ) = p
V (X)V (Y )
2 V (X)V (Y )
|| V (X)V (Y )
showing that
(X, Y ) = (X, Y ) =
which says that
(X, Y )
||
(X, X) = (X, X) = 1
(16d)
(16e)
and when Cov(X, Y ) ' 1, we say that the random variables are negatively
correlated. We also note that, in general,
1 (X, Y ) 1
18
(16f)
for all random variables X and Y . To prove Equation (16f), we note that
X
Y
X Y
Y
X
=V
+V
2Cov
,
0 V
X Y
X
Y
X Y
1
2
1
V (X) + 2 V (Y )
Cov(X, Y )
=
2
X
Y
X Y
1
1 2
+ 2 Y2 2(X, Y ) = 1 + 1 2(X, Y )
=
2 X
X
Y
and so we see that
0 1 (X, Y )
which says that 0 1 + (X, Y ), or 1 (X, Y ), and 0 1 (X, Y ), or
(X, Y ) 1, and so collectively, we have Equation (16f).
Example #9: Indicator Random Variables
Let IA and IB be indicator variables for the events A and B, which says that
if A occurs
1,
IA =
and
IB =
1,
if B occurs
Then E(IA ) = P (A) and E(IB ) = P (B) and E(IA IB ) = P (A B). This says
that
Cov(IA , IB ) = E(IA IB ) E(IA )E(IB ) = P (A B) P (A)P (B)
or
Cov(IA , IB ) = (P (A|B) P (A))P (B) = P (A)(P (B|A) P (B)).
Thus, we obtain the quite intuitive result that the indicator variables for A and
B are either positively correlated, uncorrelated, or negative correlated, depending
on whether P (A|B) is, respectively, greater than, equal to, or less than P (A).
19
Example #10: Sample Mean and Deviation From The Sample Mean
and all deviations from the sample
Let us now show that the sample mean X
X)
E(Xi X)E(
E((Xi X)
X)
X
2 ) E(Xi X)E(
X)
E(Xi X
E(X
2 ) (E(Xi ) E(X))E(
X)
E(Xi X)
2
E(X
) E(Xi )E(X)
+ (E(X))
2
E(Xi X)
(E(X
2 ) (E(X))
2)
E(Xi )E(X)
E(Xi X)
or
X)
= Cov(Xi , X)
V (X).
Cov(Xi X,
But this is
or
X
X)
= Cov Xi , 1
Cov(Xi X,
Xj
n j=1
n
1
2
n
1X
1
Cov(Xi , Xj ) 2
n j=1
n
n
X)
=
Cov(Xi X,
where 2 = V (Xi ) for each i = 1, 2, 3, ..., n. But since the Xk s are independent
and identical, we have
0, when i 6= j
Cov(Xi , Xj ) =
2
, when i = j
and so
X)
= 1 2 1 2 = 0.
Cov(Xi X,
n
n
Note that although X and Xi X are uncorrelated, they are not (in general) independent. However, in the special case where the Xi are normal random variables,
independent of a single deviation Xi X,
but it
it turns out that not only is X
for i = 1, 2, 3, ..., n.
is independent of the entire sequence of deviations Xi X
20
4. Conditional Expectation
If X and Y are jointly discrete random variables, then the conditional probability mass function of X, given that Y = y (defined for all y such that P (Y = y) > 0)
by
P ((X = x) (Y = y))
pX|Y (x|y) = P (X = x|Y = y) =
P (Y = y)
or
p(x, y)
pX|Y (x|y) =
.
(17a)
pY (y)
It is therefore natural to define, in this case, the conditional expectation of X
given that Y = y, for all values of y such that pY (y) > 0, by
X
X
xP (X = x|Y = y) =
xpX|Y (x|y).
(17b)
E(X|Y = y) =
x
Similarly, if X and Y are jointly continuous random variables with a joint probability density function f (x, y), then the conditional probability density of X, given
Y = y, is defined for all y such that fY (y) > 0, by
fX|Y (x|y) =
f (x, y)
.
fY (y)
(17c)
and the conditional expectation of X given that Y = y, for all values of y such
that fY (y) > 0, by
Z +
E(X|Y = y) =
xfX|Y (x|y)dx.
(17d)
Example #11
Suppose that the joint density of X and Y is given by
1
f (x, y) = eyx/y
y
for 0 < x and 0 < y. Let us compute E(X|Y = y). Towards this end, we first
have
Z +
Z +
1 yx/y
f (x, y)
where fY (y) =
e
f (x, y)dx =
dx = ey
fX|Y (x, y) =
fY (y)
y
0
21
so that
fX|Y (x, y) =
1 yx/y
e
y
ey
and then
E(X|Y = y) =
or E(X|Y = y) = y.
1
= ex/y
y
1
x ex/y dx,
y
g(x)pX|Y (x|y)
(18a)
g(x)fX|Y (x|y)dx
(18b)
for continuous X and Y , and it should be noted that E(X|Y ) is a random variable
whose value at Y = y is E(X|Y = y). Using this idea and Equations (18a,b), we
may say that
X
E(X) =
E(X|Y = y)pY (y)
(18c)
y
E(X) =
(18d)
(18e)
of travel. If we assume that the miner is at all times equally likely to choose any
of the three doors, determine the expected length of time until the miner reaches
safety.
To solve this, we let X denote the amount of time (in hours) until the miner
reaches safety and let Y denote the door he initially chooses. Then
E(X) = E(X|Y = 1)P (Y = 1) + E(X|Y = 2)P (Y = 2) + E(X|Y = 3)P (Y = 3)
1
1
1
=
E(X|Y = 1) + E(X|Y = 2) + E(X|Y = 3).
3
3
3
But
E(X|Y = 1) = 3
,
E(X|Y = 2) = 5 + E(X)
and E(X|Y = 3) = 7 + E(X) and so
1
1
1
E(X) = (3) + (5 + E(X)) + (7 + E(X))
3
3
3
and solving for E(X), we find that E(X) = 15 hours.
Example #13: Spending Money in a Store
Suppose that the number of people entering a small store on a given day is a
random variable with a mean of 50. Suppose further that the amount of money
spent by these customers are independent random variables having a common
mean of E(X) = $8. Finally, suppose also that the amount of money spent by
a customer is also independent of the total number of customers who enter the
store. Determine the expected amount of money spent in the store on a given
day.
To solve this, we let N denote the number of customers who enter the store
and we let Xi be the amount of money spent by customer i, then the total amount
of money spent on a given day can be expressed as
S=
N
X
Xi
and using
E(S) = E(E(S|N)).
i=1
we have
E(S|N) = E
N
X
i=1
Xi
23
N
X
i=1
E(Xi ) = NE(X),
and then
E(S) = E(E(S|N)) = E(NE(X)) = E(N)E(X).
For E(N) = 50 and E(X) = $8, we find that E(S) = $400.
Example #14: The Game of Craps
The game of craps is begun by rolling an ordinary pair of dice. If the sum of
the dice is 2, 3, or 12, the player loses. If it is 7 or 11, the player wins. If the sum
of the dice is any other number i, the player continues to roll the dice until the
sum is either 7 or i. If it is 7, the player loses and if it is i, the player wins. Let
the random variable R denote the number of rolls of the dice in a typical game of
craps. Let us determine the expected number of rolls in a typically game, which
we denote by E(R).
Towards this end, we note that if Pi is the probability that the sum of the dice
is i, then
i1
Pi = P14i =
36
for i = 2, 3, 4, 5, 6, 7 and to compute E(R), let us condition on S, the sum on the
initial roll of the dice, giving
E(R) =
12
X
E(R|S = i)Pi .
i=2
But
E(R|S = i) =
when
1,
i = 2, 3, 7, 11, 12
.
1 + 1/(Pi + P7 ), when i = 4, 5, 6, 8, 9, 10
The second of these follows from a geometric distribution since if the sum is a
value of i that does not immediately end the game, the dice will continue to be
rolled until the sum is either i or 7, and the number of rolls until this occurs is
geometric with parameter p = Pi + P7 . Therefore, we see that
E(R) =
+
+
+
E(R|S
E(R|S
E(R|S
E(R|S
or
1
E(R) = P2 + P3 + 1 +
P4
P4 + P7
1
1
P5 + 1 +
P6 + P7
+ 1+
P5 + P7
P6 + P7
1
1
P8 + 1 +
P9
+ 1+
P8 + P7
P9 + P7
1
P10 + P11 + P12
+ 1+
P10 + P7
or
1
2
1
3
E(R) =
+
+ 1+
36 36
3/36 + 6/36 36
1
6
1
4
5
+ 1+
+
+ 1+
4/36 + 6/36 36
5/36 + 6/36 36 36
5
1
4
1
+ 1+
+ 1+
5/36 + 6/36 36
4/36 + 6/36 36
3
2
1
1
+
+
+ 1+
3/36 + 6/36 36 36 36
or
557
1 839 243 1
+
+
+ =
' 3.376 rolls.
2 660 220 2
165
Next, let us determine the probability that a player wins in the game of craps.
Toward this end, we write
E(R) =
Pwins =
12
X
P (wins|S = i)Pi
i=2
or
Pwins
=
+
+
+
P (wins|S
P (wins|S
P (wins|S
P (wins|S
25
which becomes
Pwins
=
+
+
+
P (wins|S = 4)P4
P (wins|S = 5)P5 + P (wins|S = 6)P6 + P7
P (wins|S = 8)P8 + P (wins|S = 9)P9 + P (wins|S = 10)P10
P11 .
Recall from Chapter #3, the result that if there are two events E and F with
the probability that E occurs being P (E) and with the probability that F occurs
being P (F ), then the probability that E occurs before F is
P (E)
.
P (E) + P (F )
Therefore,
P (wins|S = 4) = P (a 4 occurs before a 7) =
P4
P4 + P7
with similar results for P (wins|S = 5), P (wins|S = 6), P (wins|S = 8), P (wins|S =
9), and P (wins|S = 10). Thus we find that
P4
P5
P6
Pwins =
P4 +
P5 +
P6 + P7
P4 + P7
P5 + P7
P6 + P7
P8
P9
P10
P8 +
P9 +
P10 + P11
+
P8 + P7
P9 + P7
P10 + P7
and this leads to
Pwins
4
5
6
3
4
5
3
+
+
+
=
3 + 6 36
4 + 6 36
5 + 6 36 36
5
4
4
3
3
2
5
+
+
+
+
5 + 6 36
4 + 6 36
3 + 6 36 36
which reduces to
Pwins =
or Pwins = 49.3%, which is almost even.
26
244
495
5. Conditional Variance
Just as we had defined the conditional expectation of X, given that Y = y,
we can also define the conditional variance of X, given that Y = y as
V (X|Y ) = E((X E(X|Y ))2 |Y ).
(19a)
(19b)
Note that
E(V (X|Y )) = E(E(X 2 |Y )) E((E(X|Y ))2 ) = E(X 2 ) E((E(X|Y ))2 )
and from E(E(X|Y )) = E(X), we have
V (E(X|Y )) = E((E(X|Y ))2 ) (E(E(X|Y )))2 = E((E(X|Y ))2 ) (E(X))2 .
Adding these, we arrive at
V (X) = E(V (X|Y )) + V (E(X|Y )).
(20)
the store and we let Xi be the amount of money spent by customer i, then the
total amount of money spent on a given day can be expressed as
S=
N
X
Xi
and
E(S) = E(E(S|N)).
i=1
But
E(S|N) = E
N
X
Xi
i=1
and
N
X
E(Xi ) = NE(X)
i=1
N
X
i=1
and then
Xi
N
X
V (Xi ) = NV (X),
i=1
(21)
and is defined for all real values of t and for a given distribution for X, M(t) is
unique and conversely, for a given M(t), the distribution for X is unique. Using
the fact that
X
(tX)k
tX
e =
k!
k=0
28
we see that
M(t) = E
!
X (tX)k
k=0
k!
k=0
(tX)k
k!
X
E(X k )
k=0
k!
tk
(22)
for k = 0, 1, 2, .... Thus we have M(0) = 1 and M 0 (0) = E(X) and M 00 (0) =
E(X 2 ) so that
V (X) = M 00 (0) (M 0 (0))2 .
Example #16: The Bernoulli Distribution
Here we have
p(x) =
and so
1 p, for X = 0
p,
for X = 1
M(t) = E(etX ) = e0 (1 p) + et p
or
M(t) = 1 + p(et 1).
(23a)
n x
p(x) =
p (1 p)nx
x
n
n
X
X
n tx x
n
nx
M(t) = E(e ) =
e p (1 p)
(pet )x (1 p)nx
=
x
x
x=0
x=0
tX
(23b)
M(t) = E(e ) =
X
x=1
tx
x1
e p(1 p)
p X
=
((1 p)et )x
1 p x=1
which reduces to
M(t) =
p
(1 p)et
pet
=
1 p 1 (1 p)et
1 (1 p)et
or
M(t) =
et
p
1+p
(23c)
x1 r
p(x) =
p (1 p)xr
r1
X
X
tX
tx x 1
r
xr
tx x 1
M(t) = E(e ) =
p (1 p)
pr (1 p)xr
e
=
e
r
1
r
1
x=r
x=r
which reduces to
or
t
pet
M(t) =
1 (1 p)et
r
p
M(t) =
et 1 + p
(23d)
x e
p(x) =
x!
M(t) = E(e ) =
X
etx x e
x!
x=0
which reduces to
X
(et )x e
= ee e
x!
x=0
M(t) = e(e 1) .
(23d)
M(t) = E(e ) =
1
ba
1
1
e
dx =
ba
ba
tx
which reduces to
M(t) =
etx dx
ebt eat
t(b a)
(23e)
M(t) = E(e ) =
tx
e e
dx =
31
e(t)x dx
which reduces to
t
0
for t < . Note that M(0) = 1 and M (0) = , as expected.
(23f)
M(t) =
1 2
1
f (x) = e 2 x
2
1 2
1
1
e e 2 x dx =
2
2
tx
which reduces to
1 2
etx e 2 x dx
1 2
M(t) = e 2 t
(23g)
1 2
1
f (z) = e 2 z
2
for < z < +, and X = + Z, and so
M(t) = et+ 2
2 t2
(23h)
(24b)
so that
MX+Y (t) = MX (t)MY (t) = (pet + 1 p)n (pet + 1 p)m
or
MX+Y (t) = (pet + 1 p)n+m
MX (t) = e1 (e 1)
and MY (t) = e2 (e 1)
so that
33
2 2
MX (t) = e1 t+ 2 1 t
2 2
and MY (t) = e2 t+ 2 2 t
so that
2 2
2 2
)=
tz 2
1 2
1
1
e 2 z dz =
2
2
e( 2 t)z dz.
Then
1
.
Mz2 (t) =
1 2t
MX (t) = (1 2t)n/2 .
34
(25a)
1
M(t) = /2
2 (/2)
s
1/2 t
/21
es
1
ds
1/2 t
or
1
M(t) = /2
2 (/2)(1/2 t)/2
s/21 es ds =
1
(/2)
2/2 (/2)(1/2 t)/2
or
M(t) = (1 2t)/2 .
(25b)
35