Variance Covariance Correlation coefficient

Lecture 9: Variance, Covariance,

Correlation Coefficient

Kateřina Staňková

Statistics (MAT1003)

May 2, 2012

beamer-tu-log
Variance Covariance Correlation coefficient

Outline

1 Variance
Definition
Standard Deviation
Variance of linear combination of RV

2 Covariance
Meaning & Definition
Examples

3 Correlation coefficient

book: Sections 4.2, 4.3

beamer-tu-log
Variance Covariance Correlation coefficient

And now . . .

1 Variance
Definition
Standard Deviation
Variance of linear combination of RV

2 Covariance
Meaning & Definition
Examples

3 Correlation coefficient

beamer-tu-log
Variance Covariance Correlation coefficient

Definition

Variance
Let X be an RV with µx = E(X ). Then the variance of X is
given by
V (X ) = E{(X − µX )2 }

beamer-tu-log
Variance Covariance Correlation coefficient

Definition

Variance
Let X be an RV with µx = E(X ). Then the variance of X is
given by
V (X ) = E{(X − µX )2 }
Notation: V (X ), Var(X ), σ 2 , σx2
| {z }
book

beamer-tu-log
Variance Covariance Correlation coefficient

Definition

Variance
Let X be an RV with µx = E(X ). Then the variance of X is
given by
V (X ) = E{(X − µX )2 }
Notation: V (X ), Var(X ), σ 2 , σx2
| {z }
book

Alternative formula

V (X ) = E{(X − µX )2 } = E(X 2 − 2 µX X + µ2X )

= E(X 2 ) − 2 µX E(X ) + µ2X = E(X 2 ) − 2 µ2X + µ2X
= E(X 2 ) − µ2X

beamer-tu-log
Variance Covariance Correlation coefficient

Definition

Variance
Let X be an RV with µx = E(X ). Then the variance of X is
given by
V (X ) = E{(X − µX )2 }
Notation: V (X ), Var(X ), σ 2 , σx2
| {z }
book

Alternative formula

V (X ) = E{(X − µX )2 } = E(X 2 − 2 µX X + µ2X )

= E(X 2 ) − 2 µX E(X ) + µ2X = E(X 2 ) − 2 µ2X + µ2X
= E(X 2 ) − µ2X

Observation
beamer-tu-log
Variance is always nonnegative!
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

Example 1
X : ] eyes on a die

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

Example 1
X : ] eyes on a die
 2
2 91 7 35
V (X ) = E(X ) − µ2X = − =
6 2 12

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

Example 1
X : ] eyes on a die
 2
2 91 7 35
V (X ) = E(X ) − µ2X = − =
6 2 12
q
35
Standard deviation: σX = 12

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

Example 1
X : ] eyes on a die
 2
2 91 7 35
V (X ) = E(X ) − µ2X = − =
6 2 12
q
35
Standard deviation: σX = 12

Example 2
n : a real number
beamer-tu-log
,
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

Example 1
X : ] eyes on a die
 2
2 91 7 35
V (X ) = E(X ) − µ2X = − =
6 2 12
q
35
Standard deviation: σX = 12

Example 2
n : a real number
beamer-tu-log
V (n) = E(n2 ) − µ2n = n2 − n2 = 0,
Variance Covariance Correlation coefficient

Standard Deviation

Standard deviation σX
s
Square root of V (X ), i.e., σX = + V (X ) = + E(X 2 ) − µ2X
p
| {z }
V (X )

Example 1
X : ] eyes on a die
 2
2 91 7 35
V (X ) = E(X ) − µ2X = − =
6 2 12
q
35
Standard deviation: σX = 12

Example 2
n : a real number
beamer-tu-log
V (n) = E(n2 ) − µ2n = n2 − n2 = 0, σn = 0
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

E(X ) =

E(X 2 ) =

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1
E(X ) = x g(x)dx
0

E(X 2 ) =

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4
E(X ) = x g(x)dx = x x− x dx
0 0 3 3
2
E(X ) =

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4 5
E(X ) = x g(x)dx = x x− x dx =
0 0 3 3 9
2
E(X ) =

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4 5
E(X ) = x g(x)dx = x x− x dx =
0 0 3 3 9
Z 1
2 2
E(X ) = x g(x)dx
0

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1 
10 10 4 5
E(X ) = x g(x)dx = x− x x dx =
0 0 3 3 9
Z 1 Z 1  
10 10 4
E(X 2 ) = x 2 g(x)dx = x2 x− x dx
0 0 3 3

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1 
10 10 4 5
E(X ) = x g(x)dx = x− x x dx =
0 0 3 3 9
Z 1 Z 1  
10 10 4 5
E(X 2 ) = x 2 g(x)dx = x2 x− x dx =
0 0 3 3 14

V (X ) = , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4 5
E(X ) = x g(x)dx = x− x x dx =
0 0 3 3 9
Z 1 Z 1  
10 10 4 5
E(X 2 ) = x 2 g(x)dx = x2 x− x dx =
0 0 3 3 14

V (X ) = E(X 2 ) − (E(X ))2 , σX =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4 5
E(X ) = x g(x)dx = x−x x dx =
0 0 3 3 9
Z 1 Z 1  
10 10 4 5
E(X 2 ) = x 2 g(x)dx = x2 x− x dx =
0 0 3 3 14
 2
5 5
V (X ) = E(X 2 ) − (E(X ))2 = − , σX =
14 9

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4 5
E(X ) = x g(x)dx = x−x x dx =
0 0 3 3 9
Z 1 Z 1  
10 10 4 5
E(X 2 ) = x 2 g(x)dx = x2 x− x dx =
0 0 3 3 14
 2
5 5 55
V (X ) = E(X 2 ) − (E(X ))2 = − = , σX =
14 9 1134

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

p
V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X , σX = + V (X )

Example 3
Let X be a continuous random variable with PDF
10 10 4
g(x) = x− x , 0<x <1 (0 elsewhere)
3 3

Z 1 Z 1  
10 10 4 5
E(X ) = x g(x)dx = x−x x dx =
0 0 3 3 9
Z 1 Z 1  
10 10 4 5
E(X 2 ) = x 2 g(x)dx = x2 x− x dx =
0 0 3 3 14
 2 r
5 5 55 55
V (X ) = E(X 2 ) − (E(X ))2 = − = , σX =
14 9 1134 1134

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)

,
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)

E(Y ) =

E(Y 2 ) =

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1
E(Y ) = y h(y )dy
0

E(Y 2 ) =

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1
E(Y ) = y h(y )dy = 5 y 5 dy
0 0

E(Y 2 ) =

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6
E(Y ) = y h(y )dy = 5 y 5 dy = y
0 0 6 y =0

E(Y 2 ) =

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6

E(Y 2 ) =

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1
E(Y 2 ) = y 2 h(y )dy
0

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1 Z 1
E(Y 2 ) = y 2 h(y )dy = 5 y 6 dy
0 0

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1 Z 1
5
E(Y 2 ) = y 2 h(y )dy = 5 y 6 dy =
0 0 7

V (Y ) = , σY =
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1 Z 1
5
E(Y 2 ) = y 2 h(y )dy = 5 y 6 dy =
0 0 7

V (Y ) = E(Y 2 ) − (E(Y ))2 , σY =

beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1 Z 1
5
E(Y 2 ) = y 2 h(y )dy = 5 y 6 dy =
0 0 7
 2
5 5
V (Y ) = E(Y 2 ) − (E(Y ))2 = − , σY =
7 6
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1 Z 1
5
E(Y 2 ) = y 2 h(y )dy = 5 y 6 dy =
0 0 7
 2
5 5 5
V (Y ) = E(Y 2 ) − (E(Y ))2 = − = , σY =
7 6 252
beamer-tu-log
Variance Covariance Correlation coefficient

Standard Deviation

Variance & standard deviation

V (X )= E{(X − µX )2 } = E(X 2 ) − µ2X

p
σX = + V (X )

Example 4
Let
h(y ) = 5 y 4 , 0<y <1 (0 elsewhere)
Z 1 Z 1  1
5 6 5
E(Y ) = y h(y )dy = 5 y 5 dy = y =
0 0 6 y =0 6
Z 1 Z 1
5
E(Y 2 ) = y 2 h(y )dy = 5 y 6 dy =
0 0 7
 2 r
2 5 2 5 5 5
V (Y ) = E(Y ) − (E(Y )) = − = , σY =
7 6 252 252
beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }
= E{(a X + b − (aµX + b))2 }

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }
= E{(a X + b − (aµX + b))2 }
= E{(a X − a µX )2 }

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }
= E{(a X + b − (aµX + b))2 }
= E{(a X − a µX )2 } = E{a2 (X − µX )2 }

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }
= E{(a X + b − (aµX + b))2 }
= E{(a X − a µX )2 } = E{a2 (X − µX )2 }
= a2 E{(X − µX )2 }

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }
= E{(a X + b − (aµX + b))2 }
= E{(a X − a µX )2 } = E{a2 (X − µX )2 }
= a2 E{(X − µX )2 } = a2 V (X )

beamer-tu-log
Variance Covariance Correlation coefficient

Variance of linear combination of RV

Theorem about V (a X + b)
Let X be an RV with variance V (X ). Then:
V (a X + b) = a2 V (X )
Proof:

V (a X + b) = E{(a X + b − µaX +b )2 }
= E{(a X + b − (aµX + b))2 }
= E{(a X − a µX )2 } = E{a2 (X − µX )2 }
= a2 E{(X − µX )2 } = a2 V (X )

beamer-tu-log
Variance Covariance Correlation coefficient

And now . . .

1 Variance
Definition
Standard Deviation
Variance of linear combination of RV

2 Covariance
Meaning & Definition
Examples

3 Correlation coefficient

beamer-tu-log
Variance Covariance Correlation coefficient

Meaning & Definition

What is covariance?
Dependence of realizations of 2 (or more) different RVs.

beamer-tu-log
Variance Covariance Correlation coefficient

Meaning & Definition

What is covariance?
Dependence of realizations of 2 (or more) different RVs.

Definition
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )}

Notation: Cov(X , Y ), σX ,Y
|{z}
book

beamer-tu-log
Variance Covariance Correlation coefficient

Meaning & Definition

What is covariance?
Dependence of realizations of 2 (or more) different RVs.

Definition
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )}

Notation: Cov(X , Y ), σX ,Y
|{z}
book

Remark
Covariance can be positive and negative (variance is always
nonnegative)
beamer-tu-log
Variance Covariance Correlation coefficient

Meaning & Definition

Definition
Cov(X , Y ) = E{(X − µX )(Y − µY )}
Notation: Cov(X , Y ), σX ,Y
|{z}
book

beamer-tu-log
Variance Covariance Correlation coefficient

Meaning & Definition

Definition
Cov(X , Y ) = E{(X − µX )(Y − µY )}
Notation: Cov(X , Y ), σX ,Y
|{z}
book

Alternative formula

beamer-tu-log
Variance Covariance Correlation coefficient

Meaning & Definition

Definition
Cov(X , Y ) = E{(X − µX )(Y − µY )}
Notation: Cov(X , Y ), σX ,Y
|{z}
book

Alternative formula

Cov(X , Y ) = E{(X − µX )(Y − µY )}

= E (XY − µY · X − µX · Y + µX µY )
= E(XY ) − µY · µX − µX · µY + µX µY
= E(XY ) − µX µY

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Notation: Cov(X , Y ), σX ,Y
|{z}
book

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Notation: Cov(X , Y ), σX ,Y
|{z}
book

Example 1
X , Y independent RVs

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Notation: Cov(X , Y ), σX ,Y
|{z}
book

Example 1
X , Y independent RVs

Cov(X , Y )

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Notation: Cov(X , Y ), σX ,Y
|{z}
book

Example 1
X , Y independent RVs

Cov(X , Y ) = E(XY ) − µX µY

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Notation: Cov(X , Y ), σX ,Y
|{z}
book

Example 1
X , Y independent RVs

Cov(X , Y ) = E(XY ) − µX µY = E(X ) · E(Y ) − µX · µY

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Let X and Y be RVs with µX = E(X ) and µY = E(Y ). Then

Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Notation: Cov(X , Y ), σX ,Y
|{z}
book

Example 1
X , Y independent RVs

Cov(X , Y ) = E(XY ) − µX µY = E(X ) · E(Y ) − µX · µY = 0

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Example 2
X , Y independent RVs with joint PDF

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Example 2
X , Y independent RVs with joint PDF

10x y 2 , 0 < x < y < 1


f (x, y ) = .
0, elswhere

Find Cov(X , Y ), E(X , Y )

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Example 2
X , Y independent RVs with joint PDF

10x y 2 , 0 < x < y < 1


f (x, y ) = .
0, elswhere

Find Cov(X , Y ), E(X , Y )

10 10 4
g(x) = x− x , (0 < x < 1) (0 elsewhere)
3 3
h(y ) = 5 y 4 , (0 < y < 1) (0 elsewhere)

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Example 2
X , Y independent RVs with joint PDF

10x y 2 , 0 < x < y < 1


f (x, y ) = .
0, elswhere

Find Cov(X , Y ), E(X , Y )

10 10 4
g(x) = x− x , (0 < x < 1) (0 elsewhere)
3 3
h(y ) = 5 y 4 , (0 < y < 1) (0 elsewhere)

E(X ) = 59 , E(X 2 ) = 5
14
, V (X ) = 55
1134
, E(Y ) = 56 , E(Y 2 ) = 75 , V (Y ) = 5
252

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Example 2
X , Y independent RVs with joint PDF

10x y 2 , 0 < x < y < 1


f (x, y ) = .
0, elswhere

Find Cov(X , Y ), E(X , Y )

10 10 4
g(x) = x− x , (0 < x < 1) (0 elsewhere)
3 3
h(y ) = 5 y 4 , (0 < y < 1) (0 elsewhere)

E(X ) = 59 , E(X 2 ) = 14
5 55
, V (X ) = 1134 , E(Y ) = 56 , E(Y 2 ) = 75 , V (Y ) = 5
252
R1R1 2
E(X , Y ) = 0 x x · y · 10 x y dy dx = 10 21

beamer-tu-log
Variance Covariance Correlation coefficient

Examples

Covariance
Cov(X , Y ) = E{(X − µX )(Y − µY )} = E(XY ) − µX µY

Example 2
X , Y independent RVs with joint PDF

10x y 2 , 0 < x < y < 1


f (x, y ) = .
0, elswhere

Find Cov(X , Y ), E(X , Y )

10 10 4
g(x) = x− x , (0 < x < 1) (0 elsewhere)
3 3
h(y ) = 5 y 4 , (0 < y < 1) (0 elsewhere)

E(X ) = 59 , E(X 2 ) = 14
5 55
, V (X ) = 1134 , E(Y ) = 56 , E(Y 2 ) = 75 , V (Y ) = 5
252
R1R1 2
E(X , Y ) = 0 x x · y · 10 x y dy dx = 10 21
10
Cov(X , Y ) = E(XY ) − µX µY = 21 − 59 56 = 378
5
beamer-tu-log
Variance Covariance Correlation coefficient

And now . . .

1 Variance
Definition
Standard Deviation
Variance of linear combination of RV

2 Covariance
Meaning & Definition
Examples

3 Correlation coefficient

beamer-tu-log
Variance Covariance Correlation coefficient

Definition of correlation coefficient ρ(X , Y )

The correlation coefficient of 2 RVs X and Y is defined as
follows:
cov(X , Y )
ρ(X , Y ) =
σX · σY

beamer-tu-log
Variance Covariance Correlation coefficient

Definition of correlation coefficient ρ(X , Y )

The correlation coefficient of 2 RVs X and Y is defined as
follows:
cov(X , Y )
ρ(X , Y ) =
σX · σY

Observations
−1 ≤ ρ(X , Y ) ≤ 1

beamer-tu-log
Variance Covariance Correlation coefficient

Definition of correlation coefficient ρ(X , Y )

The correlation coefficient of 2 RVs X and Y is defined as
follows:
cov(X , Y )
ρ(X , Y ) =
σX · σY

Observations
−1 ≤ ρ(X , Y ) ≤ 1
If X and Y are independent, then ρ(X , Y ) = 0.

beamer-tu-log
Variance Covariance Correlation coefficient

Definition of correlation coefficient ρ(X , Y )

The correlation coefficient of 2 RVs X and Y is defined as
follows:
cov(X , Y )
ρ(X , Y ) =
σX · σY

Observations
−1 ≤ ρ(X , Y ) ≤ 1
If X and Y are independent, then ρ(X , Y ) = 0.

In the previous example

q
5 55
p
Cov(X , Y ) =378 ,σX = V (X ) = 1134 ,
q
5
p
σY = V (Y ) = 252

beamer-tu-log
Variance Covariance Correlation coefficient

Definition of correlation coefficient ρ(X , Y )

The correlation coefficient of 2 RVs X and Y is defined as
follows:
cov(X , Y )
ρ(X , Y ) =
σX · σY

Observations
−1 ≤ ρ(X , Y ) ≤ 1
If X and Y are independent, then ρ(X , Y ) = 0.

In the previous example

q
5 55
p
Cov(X , Y ) = 378 ,σX = V (X ) = 1134 ,
q
5
p
σY = V (Y ) = 252
5
ρ(X , Y ) = q
55
378q
5
≈ 0.4264 beamer-tu-log
1134
· 252