2.

PRELIMINARIES
Laplace Transforms, Moment Generating
Functions and Characteristic Functions

2.1 Definitions

2.2 Theorems on Laplace Transforms

2.3 Operations on Laplace Transforms

2.4 Limit Theorems

2.5 Dirac Delta Function

2.6 Appendix: Complex Numbers

2.7 Appendix: Notes on Partial Fractions

31
 Laplace Transforms, Moment Generating
Functions and Characteristic Functions

2.1. Definitions:
Let ϕ(t) be defined on real line.
a. Moment Generating Function (MGF)
Z ∞
M GF = eθt ϕ(t)dt
−∞

Z ∞
which exists if |eθt ϕ(t)dt| < ∞
−∞

32
b. Characteristic Function (CF)
Z ∞
CF = eiyt ϕ(t)dt (i2 = −1).
−∞

Z ∞
which exists if |eiyt ϕ(t)|dt < ∞
−∞

Since eiyt = cos yt + i sin yt, |eiyt | = 1

Z ∞
CF exists if |ϕ(t)|dt < ∞
−∞

33
c. Laplace Transform
Let ϕ(t) be defined on [0, ∞) and assume
Z ∞
e−x0 t ϕ(t)dt < ∞
0

for some value of x0 ≥ 0.

Characteristic function of e−x0 t ϕ(t) is
Z ∞
CF = e−iyt e−x0 t ϕ(t)dt (sign of iyt could be + or -).
0
Z ∞
CF = e−(x0 +iy)t ϕ(t)dt
0
s = x0 + iy
Z ∞
CF = ϕ∗ (s) = e−st ϕ(t)dt for R(s) ≥ x0
0

ϕ∗ (s) is Laplace Transform of ϕ(t).

34
 Z ∞
ϕ∗ (s) = e−st ϕ(t)dt
0

Since ϕ(t) defined on [0, ∞), for ε ≥ 0

Z ∞ Z ∞
ϕ∗ (s) = lim e−st ϕ(t)dt = e−st ϕ(t)dt.
ε→0 ε 0+

Often a script L is used to denote a LaPlace transform; i.e.

L {ϕ(t)} = ϕ∗ (s)

Suppose f (t) is a pdf on [0, ∞),

Z ∞
f ∗ (s) = e−st f (t)dt , R(s) ≥ 0
0

Laplace transform is appropriate for non-negative r.v.

35
 Z t Z t
If the cdf is F (t) = f (x)dx = dF (x)
0 0
Z ∞
f ∗ (s) = e−st dF (t)
0

is Laplace-Stieltjes Transform.

Note: E(e−sT ) = f ∗ (s)

f ∗ (0) = 1, 0 ≤ f ∗ (s) ≤ 1

36
2.2 Theorems on Laplace Transforms (LT)

a. Uniqueness Theorem.
Distinct probability distributions have distinct Laplace Transforms

b. Continuity Theorem
For n = 1, 2, . . . , let {Fn (t)} be a sequence of cdf 0 s such that Fn → F .
Define {fn∗ (s)} as the sequence of LT such that L {f n (t)} = f ∗
n (s)
Z ∞
and define f ∗ (s) = e−st dF (t).
0

Then fn∗ (s) → f ∗ (s) and conversely if fn∗ (s) → f ∗ (s), then
Fn (t) → F (t)

37
c. Convolution Theorem
If T1 , T2 are independent, non-negative r.v. with p.d.f f1 (t), f2 (t) then
the pdf of T = T1 + T2 is
Z ∞
f (t) = f1 (τ )f2 (t − τ )dτ
0

and L {f (t)} = f1∗ (s)f2∗ (s)

In general if {Ti } i = 1, 2, . . . , n are independent non-negative r.v.,
then the Laplace transform of the pdf of T = T1 + · · · + Tn is
Yn
L {f (t)} = fi∗ (s)
i=1

38
d. Moment Generating Property
Suppose all moments exist.
∞
( )
∞ ∞
s n tn
Z Z X
∗ −st n
f (s) = e f (t)dt = (−1) f (t)dt
0 0 n=0
n!
∞ n ∞
s
X Z
= (−1)n tn f (t)dt
n=0
n! 0
∞
X sn
n
= (−1) mn where mn = E(T n )
n=0
n!

39
e. Inversion Theorem
Knowledge of f ∗ (s). The inversion formula is written
Z c+i∞
1
f (t) = est f ∗ (s)ds
2πi c−i∞
where the integration is in the complex plane and c is an appropriate
constant. (It is beyond the scope of this course to discuss the inversion
formula in detail).
Notation: L {f (t)} = f ∗ (s)
f (t) = L−1 {f ∗ (s)}
where L−1 { } is referred to as the “Inverse Laplace Transform”.

40
Example: Exponential Distribution

f (t) = λe−λt for t ≥ 0

E(T ) = 1/λ, V (t) = 1/λ2 .
Z ∞ Z ∞
f ∗ (s) = e−st λe−λt dt = λ e−(s+λ)t dt
0 0
 Z ∞
λ −(λ+s)t λ
= (λ + s)e dt =
λ+s 0 λ+s
∞  s n
λ 1 X
f ∗ (s) = = s = (−1)n
λ+s 1+ λ n=0
λ
∞ n
 
X s n!
= (−1)n
n=0
n! λn
n!
⇒ mn = E(T n ) =
λn

41
 −λt λ −1 λ
L{λe }= , L { } = λe−λt
λ+s λ+s
L{e−λt } = (λ + s)−1 , L−1 {(λ + s)−1 } = e−λt

Note:
Z c+i∞
1
est f ∗ (s)ds = f (t)
2πi c−i∞
Z c+i∞  
1 st λ
e ds = λe−λt
2πi c−i∞ λ+s
Z c+i∞
1
est (λ + s)−1 ds = e−λt
2πi c−i∞

L−1 {(λ + s)−1 } = e−λt

42
 L−1 {(λ + s)−1 } = e−λt
Differentiating w.r. to λ
L−1 {(λ + s)−2 } = te−λt

again L−1 {2(λ + s)−3 } = t2 e−λt

..
.

(n − 1) times L−1 {(n − 1)!(λ + s)−n } = tn−1 e−λt

tn−1 e−λt
 
⇒ L = (λ + s)−n , Γ(n) = (n − 1)!
Γ(n)

n−1 −λt
 
λ(λt) e
and L = λn /(λ + s)n
Γ(n)

43
 λ(λt)n−1 e−λt
Note: is gamma distribution.
Γ(n)

If {Ti } i = 1, 2, . . . , n are independent non-negative identically

distributed r.v. following an exponential distribution with parameter λ and
T = T1 + . . . + T n
n  n
Y λ
f ∗ (s) = fi∗ (s) =
1
λ+s

44
Example: Suppose T1 , T2 are independent non-negative random variables
following exponential distributions with parameters λ1 , λ2 (λ1 6= λ2 ).
What is distribution of T = T1 + T2 ?
  
λ1 λ2
f ∗ (s) = f1∗ (s)f2∗ (s) = .
λ1 + s λ2 + s

f ∗ (s) can be written as a partial fraction; i.e.

∗ A1 A2 A1 (λ2 + s) + A2 (λ1 + s)
f (s) = + =
(λ1 + s) (λ2 + s) (λ1 + s)(λ2 + s)
λ1 λ2
where A1 = , A2 = −A1
λ2 − λ 1

45
Since L−1 {(λ + s)−1 } = e−λt

f (t) = L−1 {f ∗ (s)} = A1 L−1 {(λ1 + s)−1 } + A2 L−1 {(λ2 + s)−1 }

= A1 e−λ1 t + A2 e−λ2 t
λ1 λ2  −λ1 t −λ2 t

= e −e
λ2 − λ 1

Suppose T1 , T2 , T3 are non-negative independent r.v. so that T1 is

exponential with parameter λ1 and T2 , T3 are exponential each with
parameter λ2 . Find pdf of T = T1 + T2 + T3 .
  2
∗ λ1 λ2 A1 B1 B2
f (s) = = + +
λ1 + s λ2 + s λ1 + s λ2 + s (λ2 + s)2

f (t) = A1 e−λ1 t + B1 e−λ2 t + B2 te−λ2 t

46
Homework:
Show A1 = (λ1 − λ2 )−2 λ1 λ22
−1 −2
λ1 λ22
 
B1 = − (λ1 − λ2 ) + (λ1 − λ2 )
B2 = (λ1 − λ2 )−1 λ1 λ22

In general if fi∗ (s) = (λi /λi + s) i = 1, 2, . . . , n and

Qn
f (s) = i=1 fi∗ (s), then f(t) is called the Erlangian distribution.
∗

(Erlang was a Danish telephone engineer who used this distribution to

model telephone calls).

47
 2.3 Operations on Laplace Transforms

L{e−λt } = (λ + s)−1
L{tn−1 e−λt } = Γ(n)/(λ + s)n

Setting λ = 0 in above

L{1} = 1/s

L{tn−1 } = Γ(n)/sn

48
Homework: Prove the following relationships
Z t
ϕ∗ (s)

1. L ϕ(x)dx =
0 s

2. L {ϕ0 (t)} = sϕ∗ (s) − ϕ(0+ )

3. L {ϕ(r) (t)} = sr ϕ∗ (s) − sr−1 ϕ(0+ ) − sr−2 ϕ0 (0+ )

− . . . − ϕ(0) (0+ )

Prove (1) and (2) using integration by parts

Prove (3) by successive use of (2).

49
Suppose f (t) is a pdf for a non-negative random variable and
Z t Z ∞
F (t) = f (x)dx, Q(t) = f (x)dx
0 t

Rt ϕ∗ (s)
Since L { 0
ϕ(x)dx} = s

∗
f (s)
L {F (t)} = F ∗ (s) = s

Since Q(t) = 1 − F (t)

∗
1 1 − f (s)
L {Q(t)} = Q∗ (s) = s − F ∗ (s) = s

50
 2.4 Limit Theorems

lim ϕ(t) = lim sϕ∗ (s)

t→0+ s→∞

Proof:

L {ϕ0 (t)} = sϕ∗ (s) − ϕ(0+ )

0∗
lim ϕ (s) = lim [sϕ∗ (s) − ϕ(0+ )]
s→∞ s→∞

Z ∞
But lim e−st ϕ0 (t)dt = 0
s→∞ 0

..˙ lim [sϕ∗ (s) − ϕ(0+ )] = 0

s→∞

51
 lim ϕ(t) = lim sϕ∗ (s)
t→∞ s→0

Proof:
Z ∞ Z ∞
0 −st 0
lim L {ϕ (t)} = lim e ϕ (t)dt = ϕ0 (t)dt
s→0 s→0 0 0

Z t
= lim ϕ0 (x)dx = lim [ϕ(t) − ϕ(0+ )].
t→∞ 0 t→∞

Since L {ϕ0 (t)} = sϕ∗ (s) − ϕ(0+ )

lim [sϕ∗ (s) − ϕ(0+ )] = lim [ϕ(t) − ϕ(0+ )]

s→0 t→∞

⇒ lim sϕ∗ (s) = lim ϕ(t)

s→0 t→∞

52
 2.5. Dirac Delta Function

Sometimes it is useful to use the Dirac delta function. It is a “strange”

function and has the property.

 0 for t 6= 0
δ(t) =
 ∞ t=0

Define 
 1 if t > 0
U (t) =
 0 t≤0

Let h > 0



 0 if t > 0
U (t + h) − U (t) 
1
δ(t, h) = = if − h < t ≤ 0
h 
 h
 0

if t ≤ −h

53
Define

U (t + h) − U (t)  0 for t 6= 0
δ(t) = lim δ(t, h) = lim =
h→0 h→0 h  ∞ for t = 0

Consider
Z ∞
ϕ(x)δ(t − x)dx
0

∞  
U (t − x + h) − U (t − x)
Z
= ϕ(x) lim dx
0 h→0 h
Z ∞ ∞ 
1
Z
= lim ϕ(x)U (t − x + h)dx − ϕ(x)U (t − x)dx
h→0 h 0 0

 1 t−x+h>0
Since U (t − x + h) =
 0 otherwise

54
 Z ∞
ϕ(x)δ(t − x)dx
0
(Z )
t+h Z t
1
= lim ϕ(x)dx − ϕ(x)dx
h→0 h 0 0
(Z )
t+h  
1 ϕ(t + θh)h
= lim ϕ(x)dx = lim = ϕ(t)
h→0 h t h→0 h

(by mean value theorem)

for some value of θ 0 < θ < 1.
Z ∞
..˙ ϕ(x)δ(t − x)dx = ϕ(t)
0
Z ∞
Similarly ϕ(t) = ϕ(t − x)δ(x)dx
0

55
 Z ∞ Z ∞
ϕ(t) = ϕ(x)δ(t − x)dx = ϕ(t − x)δ(x)dx
0 0

Special Cases
ϕ(x) = 1 for all x
Z ∞ Z ∞
1= δ(t − x)dx = δ(x)dx for all t
0 0
Z ∞
⇒ δ(x)dx = 1
0

Suppose ϕ(x) = e−sx . Then

Z ∞
e−sx δ(x − t)dx = e−st
0
Z ∞
If t = 0, e−sx δ(x)dx = L{δ(x)} = 1
0

56
Example 1 Consider a non-negative random variable T such that
pn = P {T = tn }. Define
∞
X
f (t) = pn δ(t − tn )
n=1
Z ∞ ∞
X Z ∞ ∞
X
f (t)dt = pn δ(t − tn )dt = pn = 1
0 n=1 0 1
∞
X Z tn X
F (tn ) = P {T ≤ tn } = pj U (tn+1 − tj ) = f (t)dt = pj
j=1 0 j≤n

Note: If tj < tn+1 ⇒ tj ≤ tn

Z ∞ ∞
X Z ∞ ∞
X
f ∗ (s) = e−st f (t)dt = pn e−st δ(t − tn )dt = pn e−stn
0 n=1 0 n=1

57
Example 2
Consider a random variable T such that p = P {T = 0} but for T > 0
Z t2
P {t1 < T ≤ t2 } = q(x)dx
t1

⇒ f (t) = pδ(t) + (1 − p)q(t)

f ∗ (s) = p + (1 − p)q ∗ (s).
Another way of formulating this problem is to define the random variable
T by 
 0 with probability p
T =
 t with pdf q(t) for t > 0

This is an example of a random variable having both a discrete and

continuous part.

58
 2.6 Appendix
ELEMENTS OF COMPLEX NUMBERS

∧ ∧
r(x, y) r
r
θ
> >
0 0

Rectangular Polar Coordinates

Coordinates r = Modulus
θ = Amplitude

x = r cos θ, y = r sin θ

59
 p
Complex no. representation: z = x + iy, r = x2 + y 2
x
θ = tan−1 y

|z| = absolute value= r i: to be determined

z = r(cos θ + i sin θ) = reiθ

Suppose zj = aj + ibj
Addition/Subtraction: z1 ± z2 ⇒ (a1 ± a2 ) + i(b1 ± b2 )

60
Multiplication:

zj = rj eiθj , z1 z2 = r1 r2 ei(θ1 +θ2 ) z1 z2 ⇒ r = r1 r2 , θ = θ 1 + θ2

z1 z2 = r1 r2 (cos(θ1 + θ2 ) + i sin(θ1 + θ2 ) = x + iy
x = r1 r2 cos(θ1 + θ2 ) = r1 r2 (cos θ1 cos θ2 − sin θ1 sin θ2 ) = a1 a2 − b1 b2
y = r1 r2 sin(θ1 + θ2 ) = r1 r2 (sin θ1 cos θ2 + cos θ1 sin θ2 ) = b1 a2 + a1 b2

⇒ z1 z2 = (a1 + ib1 )(a2 + ib2 ) = a1 a2 + i[a1 b2 + a2 b1 ] + i2 [b1 b2 ]

If i2 = −1 Re(z1 z2 ) = a1 a2 − b1 b2
i, i2 = −1, i3 = −i, i4 = 1

⇒ i4n+1 = i, i4n+2 = −1, i4n+3 = −i, i4n = 1

z n = [r(cos θ + i sin θ)]n = rn (cos nθ + i sin nθ)

61
Exponential function
∞
X θn
eθ =
0
n!
i2 θ 2 i3 θ 3
eiθ = 1 + iθ + + + ...
2 3!
θ2 θ4 θ6 θ3 θ5
= (1 − + − + . . . ) + i(θ − + ...)
2 4! 6! 3! 5!
eiθ = cos θ + i sin θ,
e−iθ = cos θ − i sin θ, cos(−θ) = cos θ; sin −θ = − sin θ
eiθ + e−iθ eiθ − e−iθ
cos θ = , sin θ =
2 2i
cos 0 = 1, sin 0 = 0

Since eiθ = cos θ + i sin θ; z = reiθ

If z = ex+iy = ex eiy : ex : modulus, ey : amplitude
z1 z2 = ex1 +iy1 · ex2 +iy2 = ex1 +x2 +i(y1 +y2 )
62
 MOMENT GENERATING AND
CHARACTERISTIC FUNCTIONS

R∞
MGF: ψ(t) = E(e ) = −∞ ety dF (y)
tY

R∞
ψ(t) exists if −∞ | ety | dF (y) ≤ ∞
Sometimes mgf does not exist.

Importance of M GF

Uniqueness Theorem: If two random variables have the same mgf , they
have the same cdf except possibly at a countable number of points having
0 probability.

63
Continuity Theorem: Let {Xn } and X have mgf {ψn (t)} and ψ(t) with
cdf 0 s Fn (x) and F (x). Then a necessary and sufficient condition for
lim Fn (X) = F (X) is that for every t, limn→∞ ψn (t) = ψ(t), where
n→∞
ψ(t) is continuous at t = 0.

Inversion Formula: Knowledge of ψ(t) enables the pdf or frequency

function to be calculated.

Convolution Theorem: If Yi are independent with mgf ψi (t), then the

Xn Yn
mgf of S = Yi is ψS (t) = ψi (t)
1 1

If ψi (t) = ψ(t) ⇒ ψS (t) = ψ(t)n

64
 MOMENT GENERATING FUNCTIONS
DO NOT ALWAYS EXIST!

For that reason one ordinarily uses a characteristic function of a

distribution rather than the mgf .

Def. The characteristic function of a random variable y is

Z ∞
ϕ(t) = E(eity ) = eity dF (y) for −∞ < t < ∞
−∞

 
Z ∞  ity 
Z ∞
ity 
|ϕ(t)| = E(e ) ≤
 e  dF (y) = dF (y) = 1
−∞ −∞
 ity 
as e  = 1.

⇒ Characteristic Functions always exist.

65
Relation between mgf and cf .

ϕ(t) = ψ(it)

X (it)n
ϕ(t) = mn ϕ(r) (0) = ir mr
n!
The characteristic function is a Fourier Transform; i.e. for any function
g(y)
Z ∞
F.T.(q(y)) = eity q(y)dy if q(y) is pdf ⇒ c.f.
−∞

66
 RELATION TO LaPLACE TRANSFORMS

Let Y be a non-negative r.v. with pdf f (y); i.e. P {Y ≥ 0} = 1. Then it is

common to use Laplace Transforms instead of characteristic functions.

Def.: s = a + ib (a > 0). Then the Laplace Transform of f (y) is

Z ∞
f ∗ (s) = e−sy f (y)dy
0

Since e−sy f (y) = e−iby e−ay f (y)

The Laplace Transform is the equivalent of taking the Fourier Transform
of e−ay f (y)

it −1
Ex. Exponential ϕ(t) = (1 − λ)
Z ∞
λ s
f ∗ (s) = e−sy λe−λy dy = = (1 + )−1
0 λ+s λ

67
 MOMENT GENERATING FUNCTIONS AND
CHARACTERISTIC FUNCTIONS

mgf cf
Bernoulli (pet + q) (peit + q)
Binomial (pet + q)n (peit + q)n
λ(et−1 ) λ(eit −1)
Poisson e e
Geometric pet /(1 − qet ) peit /(1 − qeit )
etb −eta eitb − eita
Uniform over (a, b) t(b−a) it(b − a)
tm+ 12 σ 2 t2 itm− 12 σ 2 t2
Normal e e
λ it −1
Exponential λ−t = (1 − λt )−1 (1 − )
λ

68
 INVERSION THEOREM

Integer valued R.V.

Let Y = 0, ±1, ±2, . . . with prob. f (j) = P (Y = j)
∞
X
cf ϕ(t) = E(eitY ) = eitj f (j)
−∞

Inversion Formula:
π
1
Z
f (k) = e−ikt ϕ(t)dt
2π −π

69
  
π ∞ ∞ π
1 1
Z X X Z
f (k) = e−ikt  eitj f (j) dt = f (j) ei(j−k)t dt
2π −π j=−∞
2π j=−∞ −π

Z π Z π
i(j−k)t
e dt = [cos(j − k)t + i sin(j − k)t]dt
−π −π

 π sin(j − k)t  π
 
cos(j − k)t
=  + i
 − for j 6= k

−π j − k −π (j − k)
sin nπ = 0 for any integer n.
cos nπ = cos(−nπ)

Z π  0 for j 6= k
ei(j−k)t dt =
−π  2π for j = k

70
 INVERSION FORMULA FOR CONTINUOUS
TYPE RANDOM VARIABLES

Z ∞Y have pdf f (y) and cf ϕ(t) which is integrable; i.e.

Let
|ϕ(t)| dt < ∞
−∞

Inversion Formulae:
∞
1
Z
f (y) = eiyt ϕ(t)dt
2π −∞

Ex. Normal Distribution (standard normal)

Z ∞ −y 2 /2
ity e −t2 /2
ϕ(t) = e √ dy = e
−∞ 2π

71
Replace t by −t
∞ 2
−y /2
−ity e
Z
2
e √ dy = e−t /2
−∞ 2π
Interchange symbols t and y
Z ∞ −t2 /2
−ity e −y 2 /2
e √ dt = e
−∞ 2π
√
Divide by 2π
∞ 2
−y /2
1 e
Z
2
e−ity e−t /2
dt = √
2π −∞ 2π
which is the inversion formular for N (0, 1).

72
 UNIQUENESS THEOREM

Let X be an arbitrary r.v. and let Y be N (0, 1) where X and Y are

independent. Consider Z = X + cY (c is a constant).

−c2 t2 /2
ϕZ (t) = ϕX (t)e

Note that ϕZ (t) is integrable as |ϕX (t)| ≤ 1

Z ∞
1
f (z) = e−itz ϕz (t)dt
2π −∞

73
 Z b
P (a < Z ≤ b) = FZ (b) − FZ (a) = f (z)dz
a
b ∞
1
Z Z
−it(x+cy) −c2 t2 /2
= e ϕx (t)e dtdz
a 2π −∞
"Z #
∞ b
1
Z
−itx −c2 t2 /2 −itcy
= e dx ϕx (t)e e dt
2π −∞ a

Let c → 0 ⇒ z → x and dz → dx

∞  −ibt −iat

1 e −e
Z
Fx (b) − Fx (a) = ϕx (t)dt
2π −∞ −it

⇒ The distribution function is determined by its c.f.

⇒ If two r.v.’s have same characteristic function they have the same
distribution function.

74
Examples

Double Exponential f (x) = 12 e−|x| −∞<x<∞

Z ∞ Z ∞
1 1
ϕ(t) = eitx e−|x| dx = [cos tx + i sin tx]e−|x| dx
−∞ 2 2 −∞
Since sin(−tx) = − sin tx and cos(−tx) = cos(tx)
1 ∞
Z Z ∞
ϕ(t) = (cos tx)e−|x| dx = (cos tx)e−x dx
2 −∞ 0

 ∞ e−x [t sin tx − cos tx]


1
ϕ(t) = 

2
=
0 1 + t 1 + t2
as sin 0 = 0, cos 0 = 1

Double Exponential is sometimes called Laplace’s Distribution.

75
Ex. What is c.f. of the Cauchy distribution which has
1
pdf f (y) = π(1+y 2) − ∞ < y < ∞?

Note that by the inversion formula for the double exponential

Z ∞
1 |y| 1 −ity 1
e = e dt
2 2π −∞ 1 + t2

Hence for the Cauchy distribution

Z ∞
1
ϕ(t) = eity dy = e −|t|
−∞ π(1 + y 2 )

76
 PROBLEMS

1. Let Y be any r.v.

(a) Show ϕy (t) = E[cos tY ] + iE[sin tY ]
(b) Show ϕ−y (t) = E[cos tY ] − iE[sin tY ]
(c) Show ϕ−y (t) = ϕy (−t)

2. Let Y have a symmetric distribution around 0, i.e. f (y) = f (−y)

(a) Show E(sin tY ) = 0 and ϕy (t) is real valued.
(b) Show ϕy (−t) = ϕy (t)

2
3. Let X and Y be ind. ident. dist. r.v.’s. Show ϕX−Y (t) = |ϕx (t)|

77
4. Let Yj (j = 1, 2) be independent exponential r.v.’s with
E(Yj ) = 1/λj .
Consider S = Y1 + Y2
(a) Find the c.f. of S
(b) Using (i) and the inversion theorem find the pdf of S. Hint: If
f (y) is exponential with parameter λ (E(Y ) = 1/λ), then
λ 1
R ∞ −ity λ −λy
ϕ(t) = λ−it and 2π −∞ e λ−it dt = λe by the inversion
formula.

78
 2.7 Notes on Partial Fractions
Suppose N ∗ (s) and D ∗ (s) are polynomials in s. Suppose D ∗ (s) is a
polynomial of degree k and N ∗ (s) has degree < k.
k
Y
D∗ (s) = (s − si ) (Roots are distinct)
i=1

and roots of N ∗ (s) do not coincide with D ∗(s) . Then

k
N ∗ (s) X Ai
∗
=
D (s) i=1
s − si

where constants Ai are to be determined.

Since L {e−λt } = (λ + s)−1 , e−λt = L−1 {(λ + s)−1 }
 ∗  X k
N (s)
⇒ L−1 = A i e −si t
D∗ (s) i=1

79
 k
N ∗ (s) X Ai
=
D∗ (s) i=1
s − si
Pk Q
∗
N (s) i=1 Ai j6=i (s − sj )
= Qk
D∗ (s) 1 (s − si )
k
X Y
∗
N (s) = Ai (s − sj )
i=1 j6=i

Substitute s = sr
Y
∗
N (sr ) = Ar (sr − sj )
j6=r

N ∗ (sr )
Ar = Q r = 1, 2, . . . , k
j6=r (sr − sj )

80
Since
k
Y
D∗ (s) = (s − si )
i=1

k Y
dD∗ (s) X
= D0∗ (s) = (s − sj )
ds i=1 j6=i
Y
0∗
D (sr ) = (sr − sj )
j6=r

N ∗ (sr )
Ar = 0∗ r = 1, 2, . . . , k
D (sr )

81
Example:
Let Ti be independent r.v. having pdf qi (t) = λi e−λi t (λ1 6= λ2 ) and
consider T = T1 + T2 .

qi∗ (s) = λi /(λi + s) i = 1, 2

qT∗ (s) = q1∗ (s)q2∗ (s) = λ1 λ2 /(λ1 + s)(λ2 + s)

A1 A2
= +
(λ1 + s) (λ2 + s)

N ∗ (s) = λ1 λ2 , D∗ (s) = (λ1 + s)(λ2 + s), si = −λi

N ∗ (sr )
Ar = 0∗ , D0∗ (s) = (λ1 + s) + (λ2 + s)
D (sr )
λ1 λ2 λ1 λ2
A1 = , A2 =
λ2 − λ 1 λ1 − λ 2

82
  
λ λ
1 2 1 1
qT∗ (s) = −
λ2 − λ 1 s + λ 1 s + λ2

Taking inverse transforms gives

λ1 λ2  −λ1 t −λ2 t

q(t) = e −e
λ2 − λ 1

83
 Multiple Roots

If D∗ (s) has multiple roots, the methods for finding the partial fraction
decomposition is more complicated. It is easier to use direct methods.
Example: N ∗ (s) = 1, D ∗ (s) = (a + s)2 (b + s)

N∗ 1 A1 A2 B
= = + +
D∗ (s) (a + s)2 (b + s) a + s (a + s)2 b+s

1 A1 (a + s)(b + s) + A2 (b + s) + B(a + s)2

2
=
(a + s) (b + s) (a + s)2 (b + s)

1 = A1 (a + s)(b + s) + A2 (b + s) + B(a + s)2 (1)

1 = s2 [A1 + B] + s[A1 (a + b) + A2 + B(2a)] + [abA1 + A2 b + Ba2 ]

A1 + B = 0, A1 (a + b) + A2 + 2aB = 0
A1 (ab) + A2 b + Ba2 = 1

84
The above three equations result in solutions for (A1 , A2 , B). Another
way to solve for these constants is by substituting s = −a and s = −b in
(1); i.e.
substituting s = −b in (1)
⇒ B = (a − b)−2 ⇒ A1 = −B = −(a − b)−2
substituting s = −a in (1) ⇒ A2 = (b − a)−1

A1 = −(a − b)−2 , A2 = −(a − b)−1 , B = (a − b)−2

N ∗ (s) 1 A1 A2 B
= = + +
D∗ (s) (s + a)2 (s + b) s + a (s + a)2 s+b
 ∗ 
N (s)
L−1 ∗
= −(a − b) −2 −at
e − (a − b) −1 −at
te + (a − b) −2 −bt
e
D (s)

−e−at e−bt
= 2
[1 + t(a − b)] +
(a − b) (a − b)2

85
In general if D ∗ (s) = (s − s1 )d1 (s − s2 )d2 ...(s − sk )dk the
decomposition is
N ∗ (s) A1 A2 Ad 1
= + + . . . +
D∗ (s) s − s1 (s − s1 )2 (s − s1 )d1

B1 B2 Bd 2
+ + + . . . +
s − s2 (s − s2 )2 (s − s2 )d2

+ ... . +

Z1 Z2 Zdk
+ + + . . . +
(s − sk ) (s − sk )2 (s − sk )dk

86