You are on page 1of 9

# Lecture Note 3

Multiple Random Variables (Multivariate Model)
MIT 14.30 Spring 2006
Herman Bennett

6
6.1

Multiple Random Variables
Bivariate Distribution

Many experiments deal with more than one source of uncertainty. For these cases a random vector must be deﬁned to contain the multiple random variables we are interested in. An n -dimensional random vector is a function from a sample space S into �n . In the bivariate case, n = 2.

6.1.1

Discrete Model

Let (X, Y ) be a discrete bivariate random vector. The joint pmf of the random vector (X, Y ) is the function fXY (x, y ), deﬁned by: fXY (x, y ) = P (X = x, Y = y ) and satisﬁes the following properties: f (x, y ) ≥ 0 for all pairs (x, y ). �� ii) f (x, y ) = 1.
∀x ∀y

for all x and y ;

(9)

i)

• Note that fXY (x, y ) : �2 → �.

Caution: These notes are not necessarily self-explanatory notes. They are to be used as a complement

to (and not as a substitute for) the lectures.

1

2 0 0 2 . Y ) can be completely char­ acterized by its joint pmf or its joint cdf. and compute P (X ≥ 2.05 .05 . (10) • As in the univariate case. The joint cdf of (X.y )∈A P ((X.1 .05 2 .1 0 4 0 0 . Y ) is the function F (x. Y ≤ y ) = f (x. Y ) ∈ A) = f (x. y ).2 . y ). Y ≥ 3).05 . and P (|X − Y | = 1). The probability of event A is given by: � (x. P (X = 2). (11) Example 6. the bivariate distribution of (X. y ) = P (X ≤ x.05 0 0 1 . Check the properties of f (x. f (x. y ). deﬁned by: �� X ≤x Y ≤y F (x. y ).1.Herman Bennett LN3—MIT 14.30 Spring 06 • As in the univariate case.05 . y ) 0 1 2 3 0 .1 0 3 0 . y ). in the bivariate case an event A is deﬁned as a subcollection of outcomes (x.

The joint pdf of (X. Y = y0 ) = ? • ∂ 2 F (x. y ). � XY ∞ � ∞ ii) f (x. Y ) is the function fXY (x. the bivariate distribution of (X. Y ≤ y ) = −∞ −∞ x � y fXY (u. y ). y ) ≥ 0 for all (x. then P (X = x0 . F (∞. • If (X.Herman Bennett 6. y ) = P (X ≤ x. (13) • F (−∞. Y ) ∈ A) = A fXY (x. ∞) = 1. deﬁned by: � � P ((X. Y ) is the function F (x. deﬁned by: � F (x. 3 . −∞ −∞ i) • Note again that fXY (x. • As in the continuous univariate case. v )dudv. y ) at continuous points of f (x. −∞) = 0. y ). Y ) is a continuous random vector. Y ) be a continuous bivariate random vector. y ) : �2 → �.30 Spring 06 Let (X. Y ) can be com­ pletely characterized by its joint pdf or its joint cdf. F (x. y )dxdy.y ) ∂x∂y = f (x.2 Continuous Model LN3—MIT 14.1. (12) and satisﬁes the following properties: f (x. The joint cdf of (X. for every subset A of the xy -plane. y ) = 0. y ). y )dxdy = 1.

2. f (x.30 Spring 06 Example 6. y ) = � 6xy 2 if 0 < x < 1 and 0 < y < 1 0 otherwise.6.6) and P (X + Y > 1). y ) and compute P (X ≤ 0. 4 . Check the properties of f (x. Y ≤ 0.Herman Bennett LN3—MIT 14.

y )dx (continuous model) −∞ • As with any pmf/pdf.4. fX (x) and fY (y ). y ). ﬁnd fX (x) and fY (y ). more on this later). Following Example 6. y ) and f Y (y ) = � x∈� fXY (x.2. fX (x) and fY (y ) must satisfy i) f () ≥ 0 and ii) � � / =1. Y ) from the marginal distributions.1. Let (X. y )dy and f Y (y ) = ∞ fXY (x.3. deﬁned by: fX (x) = � fX (x) = −∞ � y ∈� fXY (x. • It is not always possible to recover the joint distribution of (X. y ) (discrete model) (14) ∞ � fXY (x.30 Spring 06 6. because the marginal distributions do not contain the in­ formation about the relationship between the variables (unless they are independent. Following Example 6.Herman Bennett LN3—MIT 14. Example 6. 5 . ﬁnd fX (x) and fY (y ).2 Marginal Distribution We use the concept of marginal distribution to illustrate the fact that from a bivariate distribution it is possible to recover the univariate distribution of each of the random variables included in the random vector. The marginal pmf’s/pdf’s of X and Y are the functions fX (x) and fY (y ). Example 6. Y ) be a random vector with joint pmf/pdf fXY (x.

the conditional pmf/pdf of X given Y = y .30 Spring 06 6. Example 6. y ) f Y (y ) and f (x|y ) = fXY (x. y ) and marginal pmf’s/pdf’s fX (x) and fY (y ). Y ) be a random vector with joint pmf/pdf fXY (x. is given by: f (y |x) = P (Y = y |X = x) = (discrete model) fXY (x. y ). are now impossible outcomes (zero mass).5.1. 6 . y ) fY ( y ) (16) (continuous model) � � / =1. denoted f (y |x). denoted f (x|y ). the property / = 1 is not satisﬁed anymore and we need to rescale the probabilities of the still-possible outcomes (dividing the joint by the marginal) in order to satisfy this property.3 Conditional Distribution Let (X. y ) fX (x) and f (y |x) = fXY (x. is given by: f (x | y ) = P ( X = x | Y = y ) = (discrete model) fXY (x. ﬁnd f (y |x = 1). Following Example 6. f (x|y ) and f (y |x) must satisfy i) f () ≥ 0 and ii) • Intuition: – Knowing the value of the RV X implies that many outcomes (x. the conditional pmf/pdf of Y given X = x. while at the same time keeping constant the relative likelihood between the still-possible outcomes. that before knowing X were possible. For any x such that fX (x) > 0. � � – As a result. y ) fX (x) (15) (continuous model) For any y such that fY (y ) > 0. • As with any pmf/pdf.Herman Bennett LN3—MIT 14.

Herman Bennett LN3—MIT 14. y ) and marginal pmfs/pdfs fX (x) and fY (y ). y ) = fX (x)fY (y ) ←→ P (x. Example 6. y ) = g (x)h(y ) for all x and y .1. Y ) be a random vector with joint pmf/pdf f (x. 6. • fX. are X and Y independent? 7 .2. Following Example 6. for all x and y.2. y ) = fX (x)fY (y ). • A useful way to check independence: X and Y are independent ←→ f (x.7. ﬁnd f (y |x = 0.Y (x. The knowledge that X = x gives no additional information about Y . y ) = P (x)P (y ) (discrete model).6. Following Example 6. Following Example 6.5). are X and Y independent? Example 6.8.30 Spring 06 Example 6.4 Independence Let (X. RV X and RV Y are called independent random variables if: f (x. (17) • Note that this implies: f (y |x) = fY (y ).

y ) = i) ii) iii) iv) Check that f (x. Find f (y |x = 0.5 Wrap-up � 8yx if 0 ≤ y ≤ x ≤ 1 0 otherwise. y ) satisﬁes the properties of a joint pdf. Example 6.5).Herman Bennett LN3—MIT 14. Are X and Y independent? 8 .9. f (x.30 Spring 06 6. Find the marginal distribution of X and Y .

y ) = i) ii) iii) iv) LN3—MIT 14. 9 . f (x.5).10.Herman Bennett � Example 6. Find f (y |x = 0. Are X and Y independent? 6.6 Multivariate Distribution See attached handout. Find the marginal distribution of X and Y .30 Spring 06 cyx2 if x2 ≤ y ≤ 1 0 otherwise. Find c.