Types of Random Vector

Let (S, Σ, P ) be a probability space and let X = (X, Y ) : S −→ R2 be a random vector

with joint distribution function FX .

Notations.

• Let BRn denote the set which contains all rectangles (Cartesian product of open,
closed and semi-closed intervals) and their countable union and intersection.
• Let In be a rectangle in Rn . We will denote by BIn the set which contains all
rectangles contained in In and their countable union and intersection.
Definition 1. X is a said to be a random vector of discrete type if there exists a non-
empty finite or countable set EX ⊂ R2 such that P (X = x) > 0, for every x ∈ EX , and
P (X ∈ EX ) = 1.

The set EX is called the support of X.

The function fX : R2 → R defined by

fX (x) = P (X = x) = P (X = x, Y = y)
is called the joint probability mass function of X.
Remark 2. Let X be a random vector of discrete type with support EX , joint d.f. FX
and joint p.m.f. fX .
c c
P
(1) x∈EX fX (x) = 1. Moreover, P (X ∈ EX ) = 0 and fX (x) = 0, ∀ x ∈ EX .
(2) For any A ∈ BR2 ,
X X
P (X ∈ A) = fX (x) = fX (x)IA (x).
x∈A∩EX x∈EX

(3) For x ∈ R2 ,
X
FX (x) = P (X ∈ (−∞, x]) = fX (x).
x∈(−∞,x]∩EX

Definition 3. X is a said to be a random vector of continuous type if there exists a

nonnegative function fX : R2 → R such that
Z x1 Z x2
FX (x1 , x2 ) = fX (x, y)dydx.
−∞ −∞

The set EX = {x ∈ R2 : fX (x) > 0} is called the support of X.

The function fX is called the joint probability density function of X.
Remark 4. Let X be a random vector of continuous type with support EX , joint d.f. FX
and joint p.d.f. fX .

(1) For any x ∈ R2 , fX (x) ≥ 0, and

Z ∞Z ∞
fX (x, y)dydx = 1.
−∞ −∞

(2) For any x ∈ R2 , P (X = x) = 0. Consequently, for any countable set S ⊂ R2 ,

P (X ∈ S) = 0.
1
 (3) Let a = (a1 , a2 ), b = (b1 , b2 ) ∈ R2 such that ai < bi , i = 1, 2. Let (a, b] =
(a1 , a2 ] × (b1 , b2 ]. Then
Z b1 Z b2
P (X ∈ (a, b]) = P (a1 < X ≤ b1 , a2 < Y ≤ b2 ) = fX (x, y)dydx.
a1 a2

Theorem 5. Let X = (X, Y ) : S −→ R2 be a random vector with joint distribution

function FX .

(1) Suppose that X is of discrete type with support EX and joint p.m.f. fX . Define
Rx = {y ∈ R : (x, y) ∈ EX }, Ry = {x ∈ R : (x, y) ∈ EX }.
Then X and Y are of discrete type with support
EX = {x ∈ R : (x, y) ∈ EX for some y ∈ R}
and
EY = {y ∈ R : (x, y) ∈ EX for some x ∈ R}
respectively. The marginal p.m.f.s of X and Y are respectively given by
(P
y∈Rx fX (x, y), if x ∈ EX ,
fX (x) =
0, otherwise,
and (P
x∈Ry fX (x, y), if y ∈ EY ,
fY (y) =
0, otherwise.
(2) Suppose that X is of continuous type with support EX and joint p.d.f. fX . Then
X and Y are of continuous type with marginal p.d.f.s given by
Z ∞ Z ∞
fX (x) = fX (x, y)dy and fY (y) = fX (x, y)dx
−∞ −∞

respectively.
Example 6. Let X = (X, Y ) be a random vector with joint p.m.f.
(
cy, if (x, y) ∈ A,
fX (x, y) =
0, otherwise;
where A = {(a, b) : a, b ∈ {1, 2, . . . , n}, a ≤ b}, n ≥ 2 is a fixed integer and c is a
constant.

(1) Find the value of c.

(2) Find the marginal p.m.f.s of X and Y .
(3) Find P (X > Y ), P (X = Y ) and P (X < Y ).

Solution.
P
(1) Clearly c > 0. The support EX is A. Therefore, (x,y)∈EX fX (x, y) = 1. This
implies that c ny=1 yx=1 y = 1 or c ny=1 y 2 = 1. Thus, c = n(n+1)(2n+1)
6
P P P
.
(2) The support of X is EX = {1, 2, . . . , n} and the support of Y is EY = {1, 2, . . . , n}.
For x ∈ EX , we have Rx = {x, x + 1, . . . , n} and
n  
X X n(n + 1) (x − 1)x
fX (x, y) = c y=c − .
y∈R y=x
2 2
x
2
 The marginal p.m.f. of X is then
(
3[n(n+1)−(x−1)x]
n(n+1)(2n+1)
, if x ∈ EX ,
fX (x) =
0, otherise.
For y ∈ EY , we have Ry = {1, 2, . . . , y} and
y
X X
fX (x, y) = c y = cy 2 .
x∈Ry x=1

The marginal p.m.f. of Y is then

(
3y 2
n(n+1)(2n+1)
, if x ∈ EY ,
fY (y) =
0, otherise.
(3) Let A = {(a, b) : a > b} and B = {(a, b) : a = b}. Then
P (X > Y ) = P (X ∈ A)
X
= fX (x, y)
(x,y)∈EX ∩A

= 0.

P (X = Y ) = P (X ∈ B)
X
= fX (x, y)
(x,y)∈EX ∩B
n
X 3
=c y= .
y=1
2n + 1
2(n−1)
Therefore, P (X < Y ) = 2n+1
.
Example 7. Let X = (X, Y ) be a random vector with joint p.d.f.
(
c
, if 0 < y < x < 1, c ∈ R,
fX (x, y) = x
0, otherwise.

(1) Find the value of c.

(2) Find the marginal p.d.f.s of X and Y .
(3) Find P (X > 2Y ).

Solution.
R∞ R∞ R1Rx R1
(1) Since −∞ −∞ fX (x, y)dxdy = 1. This implies that c 0 0 x1 dydx = 1 or c 0 dx =
1 or c = 1.
(2) The marginal p.d.f. of X is given by
Z ∞
fX (x) = fX (x, y)dy
−∞
(R x
1
0 x
dy, if 0 < x < 1,
=
0, otherwise
(
1, if 0 < x < 1,
=
0, otherwise.
3
 The marginal p.d.f. of Y is given by
Z ∞
fY (y) = fX (x, y)dx
−∞
(R 1
1
y x
dx, if 0 < y < 1,
=
0, otherwise
(
− ln y, if 0 < y < 1,
=
0, otherwise.
(3) Let A = {(x, y) : x > 2y}. Then
Z ∞Z ∞
P (X > 2Y ) = fX (x, y)IA (x, y)dydx
−∞ −∞
ZZ
1
= dydx
x
0<2y<x<1
Z 1 Z x/2
1
= dydx
0 0 x
1
= .
2