The University of Hong Kong

Department of Statistics & Actuarial Science

STAT2601A Probability and Statistics I
2023-2024 First Semester

Core course for App.AI , Bioinfo., Data Sci.&Eng., Dec.Analytics, Q.Fin., Risk Mgmt. and Stat. Majors:

STAT2601A Probability and Statistics I (2023-2024 First Semester)

Tutorial 11: Transformation of Multivariate Distributions

1 Revision
Transformation of Multivariate Distributions

Let X1 , X2 , . . . , Xn be jointly distributed continuous random variables with joint probability density
function fX (x1 , x2 , . . . , xn ). Also, let Yi = gi (X1 , X2 , . . . , Xn ), i = 1, 2, . . . , n for some functions g’s
which satisfy the following conditions:

1. The transformation from X’s to Y ’s is one-one correspondence.

2. The functions gi ’s have continuous partial derivatives at all points (x1 , x2 , . . . , xn ) and the n × n
Jacobian determinant is non-zero, i.e.,

∂g1 ∂g1 ∂g1

∂x1 ∂x2
··· ∂xn
∂g2 ∂g2 ∂g2
∂x1 ∂x2
··· ∂xn
J(x1 , x2 , . . . , xn ) = .. .. ... .. 6= 0
. . .
∂gn ∂gn ∂gn
∂x1 ∂x2
··· ∂xn

at all points (x1 , x2 , . . . , xn ).

Then the joint pdf of Y1 , Y2 , . . . , Yn is given by the following formula:

fY (y1 , y2 , . . . , yn ) = fX (x1 , x2 , . . . , xn ) × |J(x1 , x2 , . . . , xn )−1 |,

where xi = hi (y1 , y2 , . . . , yn ), i = 1, 2, . . . , n.
In fact, for easier calculation, the reciprocal of the Jacobian determinant J(x1 , x2 , . . . , xn ) can
be determined as follows:
∂h1 ∂h1 ∂h1
∂y1 ∂y2
··· ∂yn
∂h2 ∂h2 ∂h2
∂y1 ∂y2
··· ∂yn
J(x1 , x2 , . . . , xn )−1 = .. .. .. .. ,
. . . .
∂hn ∂hn ∂hn
∂y1 ∂y2
··· ∂yn

where hi ’s are the inverse transformations xi = hi (y1 , y2 , . . . , yn ), i = 1, 2, . . . , n.

Some Important Transformations

1. Z = X + Y
P P
Discrete case: pZ (z) = Pr(X + Y = z) = x p(x, z − x) = y p(z − y, y)

STAT2601A Probability and Statistics I ∼ Page 1 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

Continuous case:
FZ (z) = Pr(Z ≤ z) = Pr(X + Y ≤ z)
ˆ ∞ ˆ z−x
= Pr(Y ≤ z − X) = f (x, y)dydx
−∞ −∞
ˆ ∞ ˆ z−y
= Pr(X ≤ z − Y ) = f (x, y)dxdy,
−∞ −∞
ˆ ∞ ˆ ∞
0
fZ (z) = F (z) = f (x, z − x)dx = f (z − y, y)dy.
−∞ −∞

2. Z = X − Y
P P
Discrete case: pZ (z) = Pr(X − Y = z) = x p(x, x − z) = y p(z + y, y)
Continuous case:
FZ (z) = Pr(Z ≤ z) = Pr(X − Y ≤ z)
ˆ ∞ˆ ∞
= Pr(Y ≥ X − z) = f (x, y)dydx
−∞ x−z
ˆ ∞ ˆ z+y
= Pr(X ≤ z + Y ) = f (x, y)dxdy,
−∞ −∞
ˆ ∞ ˆ ∞
0
fZ (z) = F (z) = f (x, x − z)dx = f (z + y, y)dy.
−∞ −∞

3. Z = XY
P P
Discrete case: pZ (z) = Pr(XY = z) = x p(x, x/z) = y p(y/z, y)
Continuous case:
FZ (z) = Pr(Z ≤ z) = Pr(XY ≤ z)
= Pr(Y ≤ z/X, X > 0) + Pr(Y ≥ z/X, X < 0)
ˆ ∞ ˆ z/x ˆ 0 ˆ ∞
= f (x, y)dydx + f (x, y)dydx,
0 −∞ −∞ z/x
ˆ ∞ ˆ 0
1 1
fZ (z) = F 0 (z) = f (x, z/x) dx − f (x, z/x) dx
0 x −∞ x
ˆ ∞ ˆ ∞
1 1
= f (x, z/x) dx = f (z/y, y) dy.
−∞ x −∞ y
X
4. Z = Y
P P
Discrete case: pZ (z) = Pr(X/Y = z) = x p(x, xz) = y p(yz, y)
Continuous case:
FZ (z) = Pr(Z ≤ z) = Pr(X/Y ≤ z)
= Pr(X ≤ zY, Y > 0) + Pr(X ≥ zY, Y < 0)
ˆ ∞ ˆ zy ˆ 0 ˆ ∞
= f (x, y)dxdy + f (x, y)dxdy,
0 −∞ −∞ zy
ˆ ∞ ˆ 0
fZ (z) = F 0 (z) = f (zy, y)ydy − f (zy, y)ydy
0 −∞
ˆ ∞ ˆ ∞
|x|
= f (x, x/z) 2 dx = f (zy, y)|y|dy.
−∞ z −∞

STAT2601A Probability and Statistics I ∼ Page 2 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

2 Exercises
1. Let X and Y be random variables with joint pdf
( −y
2xe
y2
, for 0 < x < y < ∞;
f (x, y) =
0, otherwise.
Let W = Y − X, Z = Y + X. Find the joint pdf of W and Z.
Solution:
Given ( (
W =Y −X X = Z−W
2
=⇒
Z =Y +X Y = Z+W
2
∂x ∂x
− 12 1 1
J −1 (x, y) = ∂w
∂y
∂z
∂y = 1
2
1 =− .
∂w ∂z 2 2 2

The joint pdf of W and Z is given by

z+w z+w
2( z−w
2
)e− 2 1 2(z − w)e− 2
fW,Z (w, z) = × − = , 0 < w < z < ∞.
( z+w
2
)2 2 (z + w)2
Note:
z−w z+w
The domain for w and z is obtained by solving 0 < x < y < ∞ =⇒ 0 < 2
< 2
< ∞.
2. Suppose X, Y and Z are independently and identically distributed as Exp(1). Derive the joint
distribution of U = X + Y , V = X + Z and W = Y + Z.
Solution:
The joint pdf of X, Y, Z is
f (x, y, z) = e−x e−y e−z
= e−(x+y+z) , x > 0, y > 0, z > 0.
Let
U = X + Y, V = X + Z, W = Y + Z
1 1 1
=⇒ X = (U + V − W ), Y = (U + W − V ), Z = (V + W − U ).
2 2 2

Hence this is 1-to-1 transformation.

∂u ∂u ∂u
∂x ∂y ∂z 1 1 0
∂v ∂v ∂v
J(x, y, z) = ∂x ∂y ∂z = 1 0 1 = −2.
∂w ∂w ∂w 0 1 1
∂x ∂y ∂z
As a result, the joint pdf of U, V, W is given by
−1 1 1
g(u, v, w) = e−(x+y+z) = e− 2 (u+v+w) , u > 0, v > 0, w > 0.
2 2

STAT2601A Probability and Statistics I ∼ Page 3 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

3. Suppose X and Y have joint density function

(
1
2 2, for x > 1, y > 1;
f (x, y) = x y
0, otherwise.

(a) Find the joint density function of U = XY and V = X/Y .

(b) Find the marginal pdfs of U and V .

Solution:

(a) Let

U = XY, V = X/Y
√
r
U
=⇒ X = U V , Y = , X > 1, Y >1
√
V √
∂x ∂x v u
√ √ −1 1 −1
J −1 (x, y) = ∂u
∂y
∂v
∂y = 2 u 2 √v
− u = − = .
√1 √ 4v 4v 2v
∂u ∂v 2 uv 2 v3

The joint pdf of U and V is given by

1 v −1 1 1
fU,V (u, v) = = 2 , ≤ v ≤ u, 1 ≤ u < ∞.
uv u 2v 2u v u

(b) The marginal pdf of U is

ˆ u
1 1 ln u
fU (u) = 2
dv = 2 [ln v]u1 = 2 , 1 ≤ u < ∞.
1
u
2u v 2u u u

As u1 ≤ v ≤ u, that is, u ≥ max v1 , v .

For 0 < v < 1, ˆ ∞  ∞

1 1 1 1
fV (v) = 2
du = − = .
1
v
2u v 2v u 1 2
v

For 1 ≤ v < ∞,
ˆ ∞  ∞
1 1 1 1
fV (v) = 2
du = − = 2.
v 2u v 2v u v 2v
Therefore, the marginal pdf of V is
(
1
2
, for 0 < v < 1;
fV (v) = 1
2v 2
, for 1 ≤ v < ∞.

STAT2601A Probability and Statistics I ∼ Page 4 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester
i.i.d.
4. Suppose X, Y ∼ U(0, 1). Find the pdf of Z = X + Y .
Solution:
The c.d.f. of Z is given by
FZ (z) = Pr(Z ≤ z) = Pr(X + Y ≤ z).
For 0 ≤ z ≤ 1,
ˆ z ˆ z−x ˆ z ˆ z−x z
x2 z2

FZ (z) = f (x, y)dydx = 1dydx = zx − = .
0 0 0 0 2 0 2
For 1 ≤ z ≤ 2,
ˆ ˆ
FZ (z) = f (x, y)dydx
x+y≤z
ˆ z−1 ˆ 1 ˆ 1 ˆ z−x
= f (x, y)dydx + f (x, y)dydx
0 0 z−1 0
ˆ z−1 ˆ 1 ˆ 1 ˆ z−x
= 1dydx + 1dydx
0 0 z−1 0
2 1
z2
 
x
= (z − 1) + zx − = 2z − − 1.
2 z−1 2
Hence, 

 0, for z < 0;
 z2 ,

for 0 ≤ z ≤ 1;
2
FZ (z) = z2


 2z − 2
− 1, for 1 < z ≤ 2;

1, for z > 2.
Then the p.d.f. of Z is 
z,
 for 0 ≤ z ≤ 1;
fZ (z) = 2 − z, for 1 < z ≤ 2;

0, otherwise.


y y
z

1 1

z region of region of
integration integration
z−1

x x
0 z 1 0 z−1 1 z
x + y = z where z < 1 x + y = z where z > 1

STAT2601A Probability and Statistics I ∼ Page 5 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

5. Let X ∼ Po(α), Y ∼ Po(β) be two independent Poisson random variables.

(a) Determine the probability mass function and moment generating function of X + Y .
(b) Determine the moment generating function of X − Y .

Solution:

(a) The pmf of X + Y is

∞
X
pX+Y (z) = Pr(X + Y = z) = Pr(X = x, X + Y = z)
x=0
∞
X
= Pr(X = x, Y = z − x)
x=0
z
X e−α αx e−β β z−x
= ∵ Pr(Y = z − x) = 0 if x > z
x=0
x! (z − x)!
z
−(α+β)
X 1
= e αx β z−x
x=0
x!(z − x)!
−(α+β) z
e X z!
= αx β z−x
z! x=0
x!(z − x)!
z  
e−(α+β) X z x z−x
= α β (then consider binomial theorem)
z! x=0
x
e−(α+β) (α + β)z
= , z = 0, 1, 2, . . . .
z!

The moment generating function of X + Y is

MX+Y (t) = E et(X+Y ) = E(etX )E(etY ) = MX (t)MY (t)

t t t
= eα(e −1) × eβ(e −1) = e(α+β)(e −1) .

Note: By the pmf or mgf of X + Y , we can see that X + Y ∼ Po(α + β).

(b) Similarly, the moment generating function of X − Y is
t −t −1)
MX−Y (t) = E(e(X−Y )t ) = E(eXt )E(eY (−t) ) = eα(e −1)+β(e .

STAT2601A Probability and Statistics I ∼ Page 6 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

6. In the past, the overall assessment of the course STAT2601 was made up by coursework (25%)
and a final examination (75%). Suppose that the coursework mark and final examination mark
of a student of STAT2601 are independently distributed as N(80, 25) and N(70, 64) respectively.
Determine the probability that this student can have an overall score higher than 85.
[How about using the current weightings? Will the probability be greater or smaller?]
Solution:
Let X and Y denote the coursework mark and final examination mark respectively. Then
they are independently distributed as X ∼ N(80, 25) and Y ∼ N(70, 64). The overall score
S = 0.25X + 0.75Y is a linear combination of X and Y and is also distributed as normal, with
mean and variance being

E(S) = 0.25 × 80 + 0.75 × 70 = 72.5, and

Var(S) = 0.252 × 25 + 0.752 × 64 = 37.5625.

Therefore S ∼ N(72.5, 37.5625) and the probability that the student can have an overall score
higher than 85 can be calculated as
 
85 − 72.5
Pr(S > 85) = 1 − Φ √ = 1 − Φ(2.04) = 1 − 0.9793 = 0.0207.
37.5625

7. Let X and Y respectively denote the weights (in kg) of a male freshman before and after a health
fitness programme. Suppose that X and Y have a bivariate normal distribution, with marginal
distributions N(75, 121) and N(72, 144), respectively; and correlation coefficient ρ = 0.9.

(a) Let Z = X − Y represent the weight loss of a random male freshman after completing the
health fitness programme. Find the distribution of Z.
(b) Use the result in part (a), or otherwise, to determine the probability that a male freshman
can successfully lose weight after completing the health fitness programme.

Solution:

(a) The weight loss Z = X − Y is a linear combination of X and Y and is also distributed as
normal, with mean and variance being

E(Z) = 75 − 72 = 3, and
Var(Z) = Var(X) + Var(Y ) − 2Cov(X, Y )
= 121 + 144 − 2 × 0.9 × 11 × 12 = 27.4.

Therefore, S ∼ N(3, 27.4).

(b) The probability that a male freshman can successfully lose weight after completing the
fitness program is given by
 
0−3
Pr(X > Y ) = Pr(Z > 0) = 1 − Φ √ = Φ(0.57) = 0.7157.
27.4

STAT2601A Probability and Statistics I ∼ Page 7 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

8. Suppose that X and Y have joint pdf

(
1
x2 y 2
, for x > 1, y > 1;
f (x, y) =
0, otherwise.

X
(a) Determine the joint pdf of W = XY and Z = .
Y
(b) Determine the marginal pdf of Z defined in part (a).

Solution:
√ p
(a) Note that W = XY and Z = X/Y ⇐⇒ X = W Z, Y = W/Z (X and Y must be
positive).
Hence, the transformation is one-to-one. The support of (W, Z) can be determined as
( (√ ( (
x>1 wz > 1 z > 1/w 1/w < z < w
⇐⇒ p ⇐⇒ ⇐⇒ .
y>1 w/z > 1 z<w w>1
The Jacobian determinant can be evaluated as
√ √
∂x ∂x z w
√ √ 1 1 1
∂w ∂z 2 w 2 √z
J= ∂y ∂y = =− − =− .
∂w ∂z
√1
2 wz
− 2√wz3 4z 4z 2z

Therefore the joint pdf of W and Z can be obtained as

√
 r 
w
fW,Z (w, z) = fX,Y wz, |J|
z
1 z 1
= × × −
wz w 2z
1 1
= 2
, < z < w, w > 1;
2w z w
and zero otherwise.
(b) Note that  
1 1
< z < w ⇐⇒ w > max ,z .
w z
For 0 < z < 1,
ˆ ∞  1/z
1 1 1 1
fZ (z) = 2
dw = = .
1/z 2w z 2z w ∞ 2
For z ≥ 1, ˆ  z
∞
1 1 1 1
fZ (z) = 2
dw = = 2.
z 2w z 2z w ∞ 2z
Therefore, 
1
2,
 for 0 < z < 1;
1
fZ (z) = , for z ≥ 1;
 2z2
0, otherwise.


STAT2601A Probability and Statistics I ∼ Page 8 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

9. [STAT1301 (08-09, 2nd) Exam Q4]

Let X and Y be two random variables with joint pdf given by
(
4e−2y , if 0 < x < y < ∞;
f (x, y) =
0, otherwise.

(a) Find the marginal pdfs of X and Y . What marginal distributions do they have?
(b) Find the correlation coefficient between X and Y .
(c) Are X and Y independent? Why?
(d) Find the conditional pdf of X given that Y = y.
(e) Let Z = X + Y . Find the distribution function of Z.

Solution:
ˆ ∞
(a) The marginal density of X is fX (x) = 4e−2y dy = 2e−2x , x > 0.
ˆ x
y
The marginal density of Y is fY (y) = 4e−2y dx = 4ye−2y , y > 0.
0
That is, Y ∼ Γ(2, 2) and X ∼ Exp(2).
(b) Note that ˆ ˆ
∞ y
3
E(XY ) = 4xye−2y dxdy = .
0 0 4
1 1 1
E(X) = , Var(X) = , E(Y ) = 1, Var(Y ) = .
2 4 2
E(XY ) − E(X)E(Y ) 1
Corr(X, Y ) = =√ .
σX σY 2

(c) No. The range of x depends on y.

f (x, y) 1
(d) The conditional pdf of X given Y = y is fX|Y (x|y) = = , 0 < x < y; and zero
fY (y) y
otherwise.
(e) The cdf of Z is
ˆ z/2 ˆ z−x
Pr(X + Y ≤ z) = 4e−2y dydx = 1 − 2e−z + e−2z = (1 − e−z )2 , z > 0;
0 x

and Pr(X + Y ≤ z) = 0 if z ≤ 0. Therefore, the cdf of Z is

(
(1 − e−z )2 , z > 0;
FZ (z) =
0, z ≤ 0.
The pdf of Z is (
2e−z − 2e−2z , z > 0;
fZ (z) =
0, z ≤ 0.

STAT2601A Probability and Statistics I ∼ Page 9 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

10. [STAT1301 (09-10, 1st) Exam Q3]

Let X ∼ N(3, 1) and Y ∼ N(0, 1) be two independent random variables.
(a) Show that the moment generating function of W = X 2 is
 
1 9t
MW (t) = √ exp , t < 1/2.
1 − 2t 1 − 2t
(b) Find the mean and variance of W defined in part (a).
(c) Let U = X + Y , V = X/Y . Find the joint pdf of U and V .
(d) Find the moment generating function of Z = XY .
(e) Suppose X and Y represent respectively the personal savings (in thousands of dollars) of
Jack and Susan in a month. Negative saving means taking on debt. What is the probability
that in total, they can save more than $40,000 in one year, assuming independence among
their savings in different months?
Solution:
(a) By definition, the mgf of W = X 2 is
ˆ ∞ ˆ ∞
tX 2 1 (x−3)2
tx2 − 2 1 9t 1−2t 3 2
E(e ) = √ e e dx = √ e 1−2t e− 2 (x− 1−2t ) dx
2π −∞ 2π −∞
ˆ ∞
1 9t 1 1 3 2 1 9t
= √ e 1−2t ·p e− 2(1/(1−2t)) (x− 1−2t ) dx = √ e 1−2t , t < 1/2.
1 − 2t 2π/(1 − 2t) −∞ 1 − 2t

(b) By direct calculation,

E(W ) = E(X 2 ) = Var(X) + [E(X)]2 = 10 and Var(W ) = MW
00
(0) − 102 = 38.
UV U
(c) Note that X = 1+V
and Y = 1+V
. The joint pdf of U and V is
∂x ∂x ∂u ∂u −1
∂u ∂v
uv u ∂x ∂y
fU,V (u, v) = fX,Y (x, y) ∂y ∂y = fX,Y ( , ) ∂v ∂v
∂u ∂v 1+v 1+v ∂x ∂y

1 (1 + v 2 )u2
  
|u| 6uv
= exp − − +9 , −∞ < u, v < ∞.
2π(1 + v)2 2 (1 + v)2 1+v

(d) The mgf of Z is

E(etZ ) = E(etXY ) = E E(etXY |X) = E [MY (tX)] ,
 
(by independence of X and Y )
t2 X 2 /2
= E(e ),
(recall mgf of standard normal)
2 1 9t2
= E(et W/2 ) = MW (t2 /2) = √ e 2(1−t2 ) , for − 1 < t < 1. (by (a))
1 − t2

(e) The required probability is

( 12 )  
X 40 − 36
Pr (Xi + Yi ) > 40 = Pr(N(36, 24) > 40) = Pr Z > √ = 0.2071.
i=1
24

STAT2601A Probability and Statistics I ∼ Page 10 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

11. [STAT1301 (10-11, 1st) Exam Q2]

A point (X, Y ) is chosen uniformly at random from within a unit circle.

y
1

x
−1 0 1

−1

The joint probability density function (pdf) of (X, Y ) is given by

(
1
, if x2 + y 2 ≤ 1;
f (x, y) = π
0, otherwise.

(a) Are X and Y independent? Why?

(b) Find the conditional pdf of Y given X = x, i.e., fY |X (y|x). What family of distributions does it
belong to?
(c) Find Cov(X, Y ).
Y
(d) Let Z = . Find the pdf of Z.
X
√
(e) Let R = X 2 + Y 2 . Find the hazard rate function of R.
(f) To choose a point uniformly at random from within the unit circle {(x, y) : x2 + y 2 ≤ 1}, we
can first randomly draw a point from within the unit square {(x, y) : −1 ≤ x ≤ 1, −1 ≤ y ≤ 1}
bounded by the dashed line and accept it if and only if it falls within the circle, i.e., R ≤ 1.
For 100 points uniformly and independently drawn at random from the square, what is the
probability that more than one-third of the points are accepted?

Solution:
(a) No. The range of X depends on Y .
(b) By definition,
ˆ √
1−x2
√
1 2 1 − x2
fX (x) = √ dy = , −1 ≤ x ≤ 1.
− 1−x2 π π
f (x, y) 1/π 1 p p
fY |X (y|x) = = √ = √ , − 1 − x2 < y < 1 − x2 .
fX (x) 2 1 − x2 /π 2 1 − x2
 √ √ 
Hence, Y |X is uniformly distributed on − 1 − x2 , 1 − x2 .
(c) Note that E(XY ) = E [E(XY |X)] = E {X [E(Y |X)]} = E(X · 0) = 0.
Also, E(Y ) = E [E(Y |X)] = 0. Then, Cov(X, Y ) = E(XY ) − E(X)E(Y ) = 0.

STAT2601A Probability and Statistics I ∼ Page 11 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong
 The University of Hong Kong
Department of Statistics & Actuarial Science
STAT2601A Probability and Statistics I
2023-2024 First Semester

(d) The cdf of Z is

 
Y
Pr(Z ≤ z) = Pr ≤z = Pr(Y ≤ zX, X > 0) + Pr(Y ≥ zX, X < 0).
X
If z > 0, then
ˆ (1+z 2 )−1/2 ˆ zx ˆ 1 ˆ √
1−x2
1 1
Pr(Z ≤ z) = √ dydx + √ dydx
0 − 1−x2 π (1+z 2 )−1/2 − 1−x2 π
ˆ −(1+z 2 )−1/2 ˆ 1−x2
√ ˆ 0 ˆ √ 1−x2
1 1
+ √ dydx + dydx
−1 − 1−x2 π −(1+z 2 )−1/2 zx π
ˆ −(1+z 2 )−1/2
1 2 p z
= + 1 − x2 dx + .
2 π −1 π(1 + z 2 )
If z < 0, then
ˆ (1+z 2 )−1/2 ˆ zx ˆ 0 ˆ √
1−x2
1 1
Pr(Z ≤ z) = √ dydx + dydx
0 − 1−x2 π −(1+z 2 )−1/2 zx π
ˆ (1+z 2 )−1/2
2 p z
= 1 − x2 dx + .
π 0 π(1 + z 2 )
d 1
By differentiating the above, we see that in both cases dz Pr(Z ≤ z) = π(1+z 2 )
. Hence,
1
fZ (z) = , −∞ < z < ∞.
π(1 + z 2 )
i.e., Cauchy distribution.
r2 π
(e) Note that Pr(R ≤ r) = Pr(X 2 + Y 2 ≤ r2 ) = = r2 , 0 < r < 1.
π

0, r ≤ 0;

Pr(R ≤ r) = r2 , 0 < r < 1;

1, r ≥ 1.


f (t) 2t
The hazard rate function of R is λ(t) = = , 0 < t < 1.
S(t) 1 − t2
(f) Note that Pr(the point falls within the circle) = π4 .
Let X be the number of points accepted. Then X ∼ B(100, π4 ). We want to find Pr(X > 100
3 ).
We may use normal approximation.
   
100 1
Pr X > = Pr X > 33
3 3
= Pr (X > 33)
( )
33.5 − 100( π4 )
= Pr Z > p
100( π4 )(1 − π4 )
= Pr(Z > −10.97071551)
= Pr(Z < 10.97071551) ≈ 1.

∼ End of Tutorial 11 ∼

STAT2601A Probability and Statistics I ∼ Page 12 ∼ K.P. WAT

Tutorial 11: Transformation of Multivariate Distributions Department of Statistics & Actuarial Science
The University of Hong Kong