Korea University, Department of Statistics

STAT232 Mathematical Statistics – Assignment 1 Due date: 09/18/2023

1. Suppose that X1 , X2 , . . . , Xn are a random sample from the exponential distribution

with mean β. Find the density function of the largest observation, Y = maxXi .

Sol) Using the cdf of Xi ’s

P (Y ≤ y) = P (maxXi ≤ y)
= P (X1 ≤ y)P (X2 ≤ y) · · · P (Xn ≤ y)
= [P (Xi ≤ y)]n
− βy n
FY (y) = [FX (y)]n = (1 − e ) , y > 0.
d n −y − y n−1
fY (y) = FY (y) = e β [1 − e β ] , y > 0.
dy β

2. Find the pdf of Z = X + Y when X and Y are i.i.d Uniform (0,1).

Sol) Z needs to be divided into ranges [0,1] and (1,2]

When 0 ≤ z ≤ 1,

FZ (z) = P (Z ≤ z) = P (X + Y ≤ z)
Z z Z z−x
= 1 dydx
0 0
Z z
= (z − x)dx
0
1
= z2
2
When 1 < z ≤ 2,

1
 FZ (z) = P (Z ≤ z) = 1 − P (Z > z)
Z 1 Z 1
=1− 1 dydx
z−1 z−x
Z 1
=1− (1 − z + x)dx
z−1
1
= − z 2 + 2z − 1
2




z, 0≤z≤1

∴ fZ (z) = 2 − z, 1 < z ≤ 2


0,

o.w

cf) You can solve the problem using the method of bivariate change of variables.

2
3. Find the pdf of Z = X + Y when the joint pdf of X and Y is as follows.

fX,Y (x, y) = 2(x + y), 0 ≤ x ≤ y ≤ 1.

Sol) Z needs to be divided into ranges [0,1] and (1,2]

When 0 ≤ z ≤ 1,

FZ (z) = P (Z ≤ z) = P (X + Y ≤ z)
Z z/2 Z z−x
= (2x + 2y) dy dx
0 x
Z z/2
= (z 2 − 4x2 ) dx
0
1
= z3
3
When 1 < z ≤ 2,

FZ (z) = P (Z ≤ z) = P (X + Y ≤ z)
Z z−1 Z 1 Z z/2 Z z−x
= (2x + 2y) dy dx + (2x + 2y) dy dx
0 x z−1 x
Z z−1 Z z/2
= (−3x3 + 2x + 1) dx + (z 2 − 4x2 ) dx
0 z−1
1 1
= − z3 + z2 −
3 3

3
 


z2, 0≤z≤1

∴ fZ (z) = −z 2 + 2z, 1 < z ≤ 2


0,

o.w

cf) You can solve the problem using the method of bivariate change of variables.

4. If you need to review the geometric and negative binomial distributions, read WMS
Sections 3.5 and 3.6 or SJ Section 2.5. The geometric distribution with parameter p is
the distribution of the number of trials to the first success when trials are independent
and each has probability p of success. The negative binomial distribution with param-
eters p and r is the distribution of the number of trials to the rth success. So it should
be true that if X1 , X2 , . . . , Xr are independent geometric rv’s all with parameter p,
then

Y = X1 + X2 + · · · + Xr

has the negative binomial distribution with parameters p and r. Use the mgf method
to show that this is true.

Sol) Since Xi ’s ∼ Geometric(p),

∞
X pet
MX (t) = etx (1 − p)x−1 p = (∵ infinite geometric series)
1 − (1 − p)et
x=1

Since Xi ’s are independent of each other,

r
Y pet
MY (t) = MXi (t) = ( )r
1 − (1 − p)et
i=1

When a random variable Z follows negative binomial with parameters r and p,

∞  
tZ z−1
X
tz
MZ (t) = E(e ) = e (1 − p)(z−r) pr
z=r
r − 1
∞  
t r
X z−1
= (pe ) {(1 − p)et }z−r
z=r
r − 1
∞  
t r
X k+r−1
= (pe ) {(1 − p)et }k (∵ k = z − r)
r−1
k=0

= (pe ) {1 − (1 − p)et }−r

t r

pet
=( )r
1 − (1 − p)et
The mgf of Z is same as the mgf of Y = X1 + X2 + · · · + Xr

4
 ∴ Y ∼ N B(r, p)

5. Suppose that X1 , X2 , . . . , Xn are a random sample from a Poisson distribution with

mean λ.

a) Show that the moment generating function of a Poisson distribution with mean
t
λ is M (t) = eλ(e −1) .
∞ tx −λ x ∞ tx x
X e e λ X e λ
E(etX ) = = e−λ
x! x!
x=0 x=0
∞
X (λet )x
= e−λ
x!
x=0
∞ t
t
X (λet )x e−λe
= e−λ eλe
x!
x=0
t
= e−λ eλe
t −1)
= eλ(e

b) Show that X1 + · · · + Xn follows Poisson with mean nλ.

Since Xi ’s are independent of each other,
r
t −1)
Y
MY (t) = MXi (t) = enλ(e
i=1

and it is the mgf of Poisson distribution with mean nλ.

6. Let X1 , X2 , . . . , Xn be independent random variables such that each Xi follows a

Gamma distribution with parameters αi and β.
(a) Show that the moment generating function of a Gamma distribution with param-
eters α and β is M (t) = (1 − βt)−α for t < 1/β.
(b) Show that X1 + · · · + Xn follows Gamma with parameters α1 + · · · + αn and β.

Sol)
a) The pdf of X ∼ Γ(α, β) is

1 −x
fX (x) = α
xα−1 e β , x > 0, α > 0, β > 0
Γ(α)β

5
 The mgf of X is
Z ∞
1 −x
MX (t) = E(etX ) = α
xα−1 e β etx dx
0 Γ(α)β
Z ∞
1 β α 1 1−βt
α−1 −( β )x
= Γ(α)( ) x e dx
Γ(α)β α 1 − βt 0
β
Γ(α)( 1−βt )α
1
= (1 − βt)−α , t <
β

b) Let Y = ni=1 Xi , Xi ∼ iid Γ(αi , β).

P

The mgf of Y is
Pn
MY (t) = E(etY ) = E(et i=1 Xi
)
n
Y
= E(etXi ), (∵ Xi ’s are independent)
i=1
n
Y 1
= (1 − βt)−αi , t<
β
i=1
Pn
= (1 − βt)− i=1 αi

By the uniqueness of mgf, Y = ni=1 Xi ∼ Gamma( ni=1 αi , β).

P P

7. A device contains three components, each of which has a lifetime in hours with the
pdf

2x −(x/10)2
f (x) = 102
e , 0 < x < ∞.

The device fails with the failure of one of the components. Assuming independent
lifetimes, what is the probability the device fails in the first hour of operation?

Sol)
Find the cdf of X.
Z x
2t −(t/10)2
FX (x) = P (X ≤ x) = e dt
0 102
Z (x/10)2
t 2
= e−y dy (∵ y = ( ) )
0 10
2
= 1 − e−(x/10) , x>0

6
 P (the device fails in the first hour) = 1 − P (X1 > 1)P (X2 > 1)P (X3 > 1)
= 1 − [1 − FX (1)]3
2 3
= 1 − (e−(0.1) )
= 0.02955

8. Let X have the pdf

2
f (x) = √2 e−x /2 , x > 0.
2π

Find the pdf of U = X 2 .

Sol) Using the change of variables.

g(x) = x2 is a one-to-one function given that the domain of g is limited to x > 0.

√ dx 1
U = X 2, X= U, = √ , u>0
du 2 u

√ dx
fU (u) = fX ( u)
du
2 u 1
= √ e− 2 √
2π 2 u
1 1 −u
=√ √ e 2
2π u
1 1
−1 − u
= 1 u
2 e 2, u>0
1
Γ( 2 )2 2

∴ U ∼ Gamma( 12 , 2) = χ2 (1)

9. When X follows Exp(θ), find the pdf of U = (2X/θ)1/2 .

Sol) Using the change of variables.

g(x) = (2x/θ)1/2 is one-to-one function given that the domain of g is limited to x > 0
only.

dx
U = (2X/θ)1/2 , X = (θ/2) U 2 , = θu, u>0
du
θ  dx
fU (u) = fX u2
2 du
1 − 1 ( θ u2 )
= e θ 2 θu
θ
u2
= u e− 2 (u > 0)

7
10. When X follows Gamma(α,β), find the pdf of U = 1/X.

Sol) Using the change of variables.

g(x) = 1/x is one-to-one function given that the domain of g is limited to x > 0 only.

1 1 dx  1 2
U= , X= , =− , u>0
X U du u

 1  dx
fU (u) = fX
u du
1  1 α−1
− β1u 1 2
 
= e
Γ(α) β α u u
1  1 α+1
− 1
= e βu , u > 0
Γ(α) β α u