L13 Random Vector

Random Vector
Let (S, Σ, P ) be a probability space. A (univariate) random variable describes a nu-

merical quantity of a typical outcome of a random experiment. In many experiments
an observation is expressed as a family of several separate numerical quantities and we
may be interested in simultaneously studying all of them together. Consider the following
example.
Example 1. Two distinguishable dice (labelled as D1 and D2 ) are thrown simultaneously.
The sample space is S = {(i, j) : i, j ∈ {1, 2, . . . , 6}}. For (i, j) ∈ S define
X1 ((i, j)) = i + j = sum of number of dots on uppermost faces of two dice
and
X2 ((i, j)) = |i−j| = absolute difference of number of dots on uppermost faces of two dice.
It may be of interest to study numerical characteristics X1 and X2 simultaneously.

These considerations lead to the study of the function X = (X1 , X2 ) : S → R
Notations.
• We denote by Rn the n-dimensional Euclidean space, i.e.,

Rn = {x = (x1 , x2 , . . . , xn ) : xi ∈ R, i = 1, 2, . . . , n}.
• For i = 1, 2, . . . , n, let Xi : S → R be any functions. Then the function X =
(X1 , X2 , . . . , Xn ) : S → Rn is defined as
X(w) = (X1 (w), X2 (w), . . . , Xn (w)), w ∈ S.
• For A ⊆ Rn ,
X −1 (A) = {w ∈ S : X(w) ∈ A}.
• For x = (x1 , x2 , . . . , xn ) ∈ Rn , we denote by (−∞, x] the n-dimensional interval
(−∞, x] = (−∞, x1 ] × (−∞, x2 ] × · · · × (−∞, xn ].
Definition 2. A function X : S −→ Rn is called an n−dimensional random vector
(RV) if X −1 ((−∞, x]) ∈ Σ, for all x ∈ Rn . That is, {w ∈ S : X1 (w) ≤ x1 , X2 (w) ≤
x2 , . . . , Xn (w) ≤ xn } ∈ Σ.
Example 3. Let A, B ⊆ S. Define X = (X1 , X2 ) : S → R2 by
(
1, if w ∈ A,
X1 (w) = IA (w) =
0, if w ∈
/ A;
and (
1, if w ∈ B,
X2 (w) = IB (w) =
0, if w ∈
/ B.
Then X is an RV if and only if A and B are events. (Prove!)
Theorem 4. Let X = (X1 , X2 , . . . , Xn ) : S → Rn be a given function. Then X is a
random vector if and only if X1 , X2 , . . . , Xn are random variables.
Proof. Exercise.
Remark 5. If S is finite or countable and Σ = P(Σ), then any function X = (X1 , X2 , . . . , Xn ) :
S → Rn is a random vector.
1
Joint Cumulative Distribution Function
Definition 6. Let X = (X1 , X2 , . . . , Xn ) : S → Rn be a random vector. The function

FX : Rn → R, defined by,
FX (x1 , x2 , . . . , xn ) = P ({w ∈ S : X1 (w) ≤ x1 , X2 (w) ≤ x2 , . . . , Xn (w) ≤ xn }), ∀ x ∈ Rn ,
is called the joint cumulative distribution function (joint c.d.f ) or the joint dis-
tribution function (d.f ) of the random vector X.
The joint distribution function of any subset of random variables X1 , X2 , . . . , Xn is
called a marginal distribution function of FX .
Remark 7. (1) As in the case of random variables, the set {w ∈ S : X1 (w) ≤
x1 , X2 (w) ≤ x2 , . . . , Xn (w) ≤ xn } will be denoted by {X1 ≤ x1 , X2 ≤ x2 , . . . , Xn ≤
xn }.
(2) In this course, we will mainly study 2- (and sometimes 3-) dimensional random
vectors.
(3) Let X = (X, Y ) : S → R2 be a random vector. The joint c.d.f. is a map FX :
R2 → R, defined by,
FX (x, y) = P ({X ≤ x, Y ≤ y}).
(4) The c.d.f. of X and Y are called a marginal c.d.f. of FX .
Proposition 8. Let X = (X, Y ) : S → R2 be a random vector with joint c.d.f. FX . Then
the marginal c.d.f. of X and Y are given by
FX (x) = lim FX (x, y) and FY (y) = lim FX (x, y)
y→∞ x→∞
Remark 9. Let (a1 , b1 ), (a2 , b2 ) ∈ R2 . Then we know that

P (a < X ≤ b) = P (X ≤ b) − P (X ≤ a) = FX (b) − FX (a).
Now,
P (a1 < X ≤ b1 , a2 < Y ≤ b2 )
= P (a1 < X ≤ b1 , Y ≤ b2 ) − P (a1 < X ≤ b1 , Y ≤ a2 )
= [P (X ≤ b1 , Y ≤ b2 ) − P (X ≤ a1 , Y ≤ b2 )]
− [P (X ≤ b1 , Y ≤ a2 ) − P (X ≤ a1 , Y ≤ a2 )]
= FX (b1 , b2 ) − FX (a1 , b2 ) − FX (b1 , a2 ) + FX (a1 , a2 ).
Theorem 10. Let FX be the joint cumulative distribution function of a random vector
X = (X, Y ). Then
(1) x→∞
lim FX (x, y) = 1.
y→∞
(2) limy→−∞ FX (x, y) = 0 and limx→−∞ FX (x, y) = 0.

(3) FX (x, y) is right continuous and nondecreasing in each argument (keeping other
argument fixed).
(4) For each (a1 , b1 ] × (a2 , b2 ] in R2 ,
∆ = FX (b1 , b2 ) − FX (a1 , b2 ) − FX (b1 , a2 ) + FX (a1 , a2 ) ≥ 0.
Theorem 11. Let G : R2 → R be a function which satisfies properties (1) − (4) of
Theorem 10. Then there exists a probability space (S, Σ, P ) and a random vector X =
(X1 , X2 , . . . , Xn ) defined on (S, Σ, P ) such that G is the distribution function of X.
2
Example 12. Let G : R2 → R be defined by


 x, if 0 ≤ x < 1, y ≥ 1,
y 2 , if x ≥ 1, 0 ≤ y < 1,

G(x, y) =

 1, if x ≥ 1, y ≥ 1,

0, otherwise.

Show that G is not a distribution function of any random vector (X, Y ).
Solution. Clearly G satisfies properties (1) − (3) of Theorem 10.
For (a1 , b1 ] × (a2 , b2 ], where a1 , a2 ∈ [0, 1), b1 , b2 ∈ [1, ∞) and a1 + a22 > 1. Then
G(b1 , b2 ) − G(a1 , b2 ) − G(b1 , a2 ) + G(a1 , a2 ) = 1 − a1 − a22 + 0 < 0.
Thus, G is not a joint c.d.f. of any random vector.
Example 13. Consider the function G : R2 → R defined by
 2

 xy , if 0 ≤ x < 1, 0 ≤ y < 1,
x, if 0 ≤ x < 1, y ≥ 1,



G(x, y) = y 2 , if x ≥ 1, 0 ≤ y < 1,

1, if x ≥ 1, y ≥ 1,




0, otherwise.
(1) Show that G is a joint c.d.f. of some random vector (X, Y ).

(2) Find the marginal c.d.f. of X and Y .
lim G(x, y) = 1. For fixed x ∈ R, limy→−∞ G(x, y) = 0 and for fixed

Solution. Clearly x→∞
y→∞
y ∈ R, limx→−∞ G( x, y) = 0.
We note that if y < 0, then G(x, y) = 0 for all x ∈ R. Moreover,

0, if x < 0,

G(x, y) = xy 2 , if 0 ≤ x < 1, 0 ≤ y < 1,
y 2 , if x ≥ 1,

and 
0, if x < 0,

G(x, y) = x, if 0 ≤ x < 1, y ≥ 1,

1, if x ≥ 1.
One can see that for y ∈ R, G(x, y) is a continuous (and hence right continuous) function
of x. Similarly, for each x ∈ R, G(x, y) is a continuous function of y
Furthermore, G(x, y) is non-decreasing in each argument keeping other argument fixed.
For (a1 , b1 ] × (a2 , b2 ], we need to show that ∆ = G(b1 , b2 ) − G(a1 , b2 ) − G(b1 , a2 ) +

G(a1 , a2 ) ≥ 0. We consider the following cases.
(1) a1 < 0. Then ∆ = G(b1 , b2 ) − G(b1 , a2 ) ≥ 0 as G is nondecreasing.

(2) a2 < 0.
(3) 0 ≤ a1 < 1, 0 ≤ a2 < 1, 0 ≤ b1 < 1, 0 ≤ b2 < 1.
(4) 0 ≤ a1 < 1, 0 ≤ a2 < 1, 0 ≤ b1 < 1, b2 ≥ 1.
(5) 0 ≤ a1 < 1, 0 ≤ a2 < 1, b1 ≥ 1, 0 ≤ b2 < 1.
(6) 0 ≤ a1 < 1, 0 ≤ a2 < 1, b1 ≥ 1, b2 ≥ 1.
3
(7) 0 ≤ a1 < 1, a2 ≥ 1, 0 ≤ b1 < 1, b2 ≥ 1.
(8) 0 ≤ a1 < 1, a2 ≥ 1, b1 ≥ 1, b2 ≥ 1.
(9) a1 ≥ 1, 0 ≤ a2 < 1, b1 ≥ 1, 0 ≤ b2 < 1.
(10) a1 ≥ 1, 0 ≤ a2 < 1, b1 ≥ 1, b2 ≥ 1.
(11) a1 ≥ 1, a2 ≥ 1, b1 ≥ 1, b2 ≥ 1.
In all these cases verify that ∆ ≥ 0.

Therefore, G(x, y) is a distribution function of some random vector (X, Y ).
The marginal c.d.f. of X and Y are respectively

0, if x < 0,

FX (x) = lim G(x, y) = x, if 0 ≤ x < 1,
y→∞ 
1, if x ≥ 1,
and 
0, if y < 0,

FY (y) = lim G(x, y) = y 2 , if 0 ≤ y < 1,
x→∞ 
1, if y ≥ 1.

L13 Random Vector

Uploaded by

Copyright:

Available Formats

L13 Random Vector

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

L13 Random Vector

Uploaded by

Copyright:

Available Formats

Random Vector

Let (S, Σ, P ) be a probability space. A (univariate) random variable describes a nu-

It may be of interest to study numerical characteristics X1 and X2 simultaneously.

• We denote by Rn the n-dimensional Euclidean space, i.e.,

Definition 6. Let X = (X1 , X2 , . . . , Xn ) : S → Rn be a random vector. The function

Remark 9. Let (a1 , b1 ), (a2 , b2 ) ∈ R2 . Then we know that

(2) limy→−∞ FX (x, y) = 0 and limx→−∞ FX (x, y) = 0.

Show that G is not a distribution function of any random vector (X, Y ).

Solution. Clearly G satisfies properties (1) − (3) of Theorem 10.

(1) Show that G is a joint c.d.f. of some random vector (X, Y ).

lim G(x, y) = 1. For fixed x ∈ R, limy→−∞ G(x, y) = 0 and for fixed

For (a1 , b1 ] × (a2 , b2 ], we need to show that ∆ = G(b1 , b2 ) − G(a1 , b2 ) − G(b1 , a2 ) +

(1) a1 < 0. Then ∆ = G(b1 , b2 ) − G(b1 , a2 ) ≥ 0 as G is nondecreasing.

In all these cases verify that ∆ ≥ 0.

You might also like