STAT/MATH 395 A - PROBABILITY II – UW

Winter Quarter 2017 Néhémy Lim

HW3 : Moment functions – Solutions

Problem 1. Let X be a real-valued random variable on a probability space

(Ω, A, P) with moment generating function MX .
(a) Show that for any constants a, b ∈ R,

MaX+b (t) = ebt MX (at)

Answer. For a given t in the domain of MX , we have :

h i
MaX+b (t) = E et(aX+b)
= etb E eta X
 

= etb MX (ta)

(b) Application. Let X follow an exponential distribution with parameter λ > 0.

Compute the moment generating function of Y = aX (a > 0), and conclude
that Y follows an exponential distribution. Specify the new parameter.
Answer. For a given t, we have

MY (t) = MaX (t)

= et·0 MX (ta)
λ
=
λ − at
λ/a
=
λ/a − t

We recognize the moment generating function of an exponential distribution

with parameter λ/a. By virtue of the uniqueness property of moment gener-
ating functions, this proves that Y = aX follows an exponential distribution
with parameter λ/a.

Problem 2. Let X be a continuous random variable with probability density

function :
fX (x) = xe−x 1[0,∞) (x)

1
(a) Show that MX (t) the moment generating function of X is given by :
1
MX (t) =
(1 − t)2

Answer.
Z ∞
tX
MX (t) = E[e ]= etx fX (x) dx
−∞
Z ∞
= etx xe−x dx
0
Z ∞
= xe−(1−t)x dx
0

The following change of variable : u = (1 − t)x, du = (1 − t) dx gives : for

t < 1,
Z ∞
u −u du
MX (t) = e
0 1 − t 1−t
Z ∞
1
= ue−u du
(1 − t)2
| 0 {z }
=1 since fX is a pdf
1
=
(1 − t)2

(b) Compute E[X] using MX (t).

Answer. Let us differentiate MX (t) : for t < 1,

0 2
MX (t) =
(1 − t)3
Thus,
0 2
E[X] = MX (0) = =2
(1 − 0)3
(c) Compute Var(X) using MX (t).
00
Answer. The second derivative MX (t) is given by : for t < 1,

00 6
MX (t) =
(1 − t)4
Therefore, the second moment of X is :

00 6
E[X 2 ] = MX (0) = =6
(1 − 0)4
Finally,
00 0
Var(X) = MX (0) − MX (0)2 = 6 − 22 = 2.

2
Problem 3. Let X be a continuous random variable with probability density
function fX and moment generating function MX defined on a neighborhood
(−h, h) of zero, for some h > 0. Show that1

P(X ≥ a) ≤ e−at MX (t) for 0 < t < h

Hint. Start from the right-hand side of the inequality and split the integral
defining MX (t) into the intervals (−∞, a) and (a, ∞).
Answer.
Z ∞
MX (t) = E[etX ] = etx fX (x) dx
−∞
Z a Z ∞
= etx fX (x) dx + etx fX (x) dx
−∞ a
Ra
Noting that −∞
etx fX (x) dx ≥ 0 since etx ≥ 0 and fX (x) ≥ 0, we have
Z ∞
MX (t) ≥ etx fX (x) dx
a
Z ∞
≥ eat fX (x) dx
a
tx at
since e ≥e for all x ≥ a. Therefore,
Z ∞
at
MX (t) ≥ e fX (x) dx = eat P(X ≥ a).
a

which completes the proof.

Problem 4. Let X be a real-valued random variable on a probability space

(Ω, A, P) with moment generating function MX . The cumulant generating
function of X, denoted KX , is defined by

KX (t) = ln MX (t)

(a) Show that

0
KX (0) = E[X]

Answer.
0
0 MX (t)
KX (t) =
MX (t)
Thus,
0
0 MX (0)
KX (0) = = E[X]
MX (0)
0
since MX (0) = E[X] and MX (0) = 1.
1 Actually, this property also holds for discrete random variables.

3
(b) and that
00
KX (0) = Var(X)

Answer.
00 0
00 MX (t)MX (t) − MX (t)2
KX (t) =
MX (t)2
Hence,
00 0
00 MX (0)MX (0) − MX (0)2
KX (0) = 2
MX (0)
E[X 2 ] · 1 − (E[X])2
=
12
= Var(X)

(c) Application. We say that a random variable X follows a geometric distribu-

tion G(p), with parameter p ∈ (0, 1) if its probability mass function is given
by :
pX (x) = P(X = x) = (1 − p)x−1 p, for x ∈ N∗
Compute MX (t) the moment generating function of X.
Answer.
∞
X
MX (t) = etx pX (x)
x=1
X∞
= etx (1 − p)x−1 p
x=1
∞
X x−1
= pet (1 − p)et

x=1

1−1
We recognize a geometric series with first term [(1 − p)et ] = 1 and com-
mon ratio (1 − p)et . This series converges if and only if −1 < (1 − p)et < 1,
that is if and only if t < − ln(1 − p). In this case, the moment generating
function is defined and is given by :

pet
MX (t) =
1 − (1 − p)et

(d) Compute KX (t) the cumulant generating function of X.

Answer.
pet
KX (t) = ln
1 − (1 − p)et
= ln p + t − ln 1 − (1 − p)et

4
(e) Compute E[X] using KX (t).
0
Answer. The first derivative KX (t) is given by :

0 (1 − p)et
KX (t) = 1 +
1 − (1 − p)et

Therefore,
0
E[X] = KX (0)
(1 − p) · 1
=1+
1 − (1 − p) · 1
1
=
p

(f) Compute Var(X) using KX (t).

00
Answer. The second derivative KX (t) is given by :

00 (1 − p)et (1 − (1 − p)et ) − (1 − p)et (−(1 − p)et )

KX (t) = 2
(1 − (1 − p)et )
(1 − p)et
= 2
(1 − (1 − p)et )

Therefore,
00
Var(X) = KX (0)
(1 − p) · 1
= 2
(1 − (1 − p) · 1)
1−p
=
p2