• Aucun résultat trouvé

80-646-08StochasticcalculusGeneviŁveGauthier Expectationandconditionalexpectation

N/A
N/A
Protected

Academic year: 2022

Partager "80-646-08StochasticcalculusGeneviŁveGauthier Expectationandconditionalexpectation"

Copied!
77
0
0

Texte intégral

(1)

Independence Conditional probability Expectation Moments Conditional expectation

Expectation and conditional expectation

80-646-08 Stochastic calculus

Geneviève Gauthier

HEC Montréal

(2)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of two events

De…nition

De…nition

Let ( , F , P ) be a probability space. Two events A and B are said to be independent if

P ( A \ B ) = P ( A ) P ( B ) .

(3)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of two random variables

De…nition

De…nition

The random variables X and Y are both built on the probability space ( Ω , F , P ) , Card ( Ω ) < ∞ . Let P

X

= f A

1

, ..., A

m

g and P

Y

= f B

1

, ..., B

n

g be two …nite partitions that respectively generate the sigma-algebras σ ( X ) and σ ( Y ) . The random variables X and Y are said to be independent if

8 A 2 P

X

and 8 B 2 P

Y

, P ( A \ B ) = P ( A ) P ( B ) .

Intuitively, X and Y are independent when having some

information about one of them doesn’t provide us with

any information about the other.

(4)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of two random variables I

Example

Example.

ω X Y P P

1 1 1

16

0

2 1 0

16 15

3 1 0

16 15

ω X Y P P

4 0 1

16 15

5 0 0

16 15

6 0 0

16 15

If A = n 1 , 2 , 3 o

and B = n 1 , 4 o

then

P

X

= f A, A

c

g and P

Y

= f B, B

c

g .

(5)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of two random variables II

Example

The random variables X and Y are independent on the probability space ( Ω , F , P ) since

P

(

A

)

P

(

B

) =

1 2

1 3

=

1

6

=

Pn 1o

=

P

(

A

\

B

)

, P

(

Ac

)

P

(

B

) =

1

2 1 3

=

1

6

=

Pn 4o

=

P

(

Ac

\

B

)

, P

(

A

)

P

(

Bc

) =

1

2 2 3

=

1

3

=

Pn 2, 3o

=

P

(

A

\

Bc

)

, P

(

Ac

)

P

(

Bc

) =

1

2 2 3

=

1

3

=

Pn 5, 6o

=

P

(

Ac

\

Bc

)

.

Intuitively, the answer to the question a dice is rolled; what

is the probability that the random variable X takes the

value of 1? is the same as the answer to the question a

dice is rolled and the random variable Y takes the value of

1; what is the probability that the random variable X also

takes the value of 1?. The answer is one half.

(6)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of two random variables III

Example

By contrast, the same random variables X and Y are dependent on the probability space ( Ω , F , P ) since

P

(

A

)

P

(

B

) =

2 5

1 5

=

2

25

6 =

0

=

P n 1o

=

P

(

A

\

B

)

.

Intuitively, the answer to the question a dice is rolled; what is the probability that the random variable X takes the value of 1? is not the same as the answer to the question a dice is rolled and the random variable Y takes the value of 1; what is the probability that the random variable X also takes the value of 1?. In the …rst case, the answer is

2

5

while in the second case, the answer is 0.

(7)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of two random variables

Remark

Remark. Although random variables may very well exist with

no need at all for a probability space (a measurable space is

su¢ cient), it is necessary to know the probability measure

associated with the measurable space to be able to speak of

independence.

(8)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of random variables I

Example

Example.

ω X Y Z P

ω

1

1 1 0

18

ω

2

1 0 0

18

ω

3

1 0 1

18

ω

4

1 0 1

18

ω X Y Z P

ω

5

0 1 1

18

ω

6

0 0 0

18

ω

7

0 0 0

18

ω

8

0 0 1

18

(9)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of random variables II

Example

Such random variables are pairwise independent since

PfX=0gPfY=0g = 1 2

3 4=3

8=Pfω678g=P(fX=0g \ fY=0g), PfX=1gPfY=0g = 1

2 3 4=3

8=Pfω234g=P(fX=1g \ fY=0g), PfX=0gPfY=1g = 1

2 1 4=1

8=Pfω5g=P(fX=0g \ fY=1g), PfX=1gPfY=1g = 1

2 1 4=1

8=Pfω1g=P(fX=1g \ fY=1g).

(10)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of random variables III

Example

PfX=0gPfZ=0g = 1 2

1 2=1

4=Pfω6,ω7g=P(fX=0g \ fZ=0g), PfX=1gPfZ=0g = 1

2 1 2=1

4=Pfω1,ω2g=P(fX=1g \ fZ=0g), PfX=0gPfZ=1g = 1

2 1 2=1

4=Pfω58g=P(fX=0g \ fZ=1g), PfX=1gPfZ=1g = 1

2 1 2=1

4=Pfω34g=P(fX=1g \ fZ=1g).

(11)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of random variables IV

Example

PfZ=0gPfY=0g = 1 2

3 4=3

8=Pfω2,ω6,ω7g=P(fZ=0g \ fY=0g), PfZ=1gPfY=0g = 1

2 3 4=3

8=Pfω348g=P(fZ=1g \ fY=0g), PfZ=0gPfY=1g = 1

2 1 4=1

8=Pfω1g=P(fZ=0g \ fY=1g), PfZ=1gPfY=1g = 1

2 1 4=1

8=Pfω5g=P(fZ=1g \ fY=1g).

(12)

Independence Events Random variables Example Conditional probability Expectation Moments Conditional expectation

Independence of random variables V

Example

But such variables are not mutually independent since P f X = 1 g P f Y = 1 g P f Z = 1 g

= 1

2 1 4

1 2

= 1

16

6

= 0

= P f?g

= P ( f X = 1 g \ f Y = 1 g \ f Z = 1 g ) .

(13)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability

De…nition

De…nition

Let ( , F , P ) be a probability space such that Card ( ) < . For all event A 2 F with a positive probability, P ( A ) > 0, the conditional probability given A, denoted P ( j A ) , is de…ned as

8 B 2 F , P ( B j A ) = P ( B \ A )

P ( A ) .

(14)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability I

Interprétation

The random experiment consists of rolling a dice.

If the dice is well balanced, what is the probability to obtain more than three points?

The correct answer is:

12

.

Now, if after rolling the dice, I inform you that the number

of points obtained is even, what is the probability that the

upper side of the dice shows more than points?

(15)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability II

Interprétation

The answer previously given should change in order to take advantage of the information provided. Since, out of three cases where the number of points is even, there are two cases where the number of points is also greater than three, the right answer is

23

.

P n

4 , 5 , 6 o n

2 , 4 , 6 o

= P

n 4 , 5 , 6 o

\ n 2 , 4 , 6 o P n

2 , 4 , 6 o

= P

n 4 , 6 o P n

2 , 4 , 6 o =

2 6 3 6

= 2

3 .

(16)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability III

Interprétation

Why do we require that the probability associated with the event on which we condition be positive?

Mathematically, it’s simply to prevent from making the silly mistake of dividing by zero.

Intuitively, if, after the dice roll, I inform you that the result is negative, the more polite among you will exclaim

”You’re such a joker!” while some other may call me a liar.

(17)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability I

Example

Example. Let’s go back to the stochastic process representing the share price of a security, to which we will add a probability measure P on the measurable space ( Ω , F ) . Recall that the sigma-algebra here is F = σ ff ω

1

, ω

2

g , f ω

3

g , f ω

4

gg .

ω X

0

( ω ) X

1

( ω ) X

2

( ω ) X

3

( ω ) P ( ω )

ω

1

1

12

1

12

?

ω

2

1

12

1

12

?

ω

3

1 2 1 1

38

ω

4

1 2 2 2

28

(18)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability II

Example

Question. Knowing that the share price is worth one dollar at

time t = 2, do the probabilities associated with the possible

share prices at time t = 3 change?

(19)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability III

Example

Answer. Yes. Let

A = f ω 2 : X

2

( ω ) = 1 g = f ω

1

, ω

2

, ω

3

g .

Since

P

(

A

) =

P

f

ω1,ω2,ω3

g =

P

f

ω1,ω2

g +

P

f

ω3

g =

3 8

+

3

8

=

3 4 alors

P X3

=

1

2

jf

X2

=

1

g

=

P

( f

ω1,ω2

g \ f

ω1,ω2,ω3

g )

P

f

ω1,ω2,ω3

g =

P

f

ω1,ω2

g

P

(

A

) =

3

8 4 3

=

1

2

6

=

3

8

=

P X3

=

1 2 ;

(20)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability IV

Example

P

( f

X3

=

1

g jf

X2

=

1

g )

=

P

( f

ω3

g \ f

ω1,ω2,ω3

g )

P

f

ω1,ω2,ω3

g =

P

f

ω3

g

P

(

A

) =

3

8 4 3

=

1

2

6

=

3

8

=

P

f

X3

=

1

g

;

P

( f

X3

=

2

g jf

X2

=

1

g )

=

P

( f

ω4

g \ f

ω1,ω2,ω3

g )

P

f

ω1,ω2,ω3

g =

P

( ? )

P

(

A

) =

0

6 =

2

8

=

P

f

X3

=

2

g

.

(21)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability V

Example

ω X0

(

ω

)

X1

(

ω

)

X2

(

ω

)

X3

(

ω

)

P

(

ω

)

ω1 1 12 1 12 ?

ω2 1 12 1 12 ?

ω3 1 2 1 1 38

ω4 1 2 2 2 28

We had determined that F = σ ff ω

1

, ω

2

g , f ω

3

g , f ω

4

gg . Then

F

A

= σ ff ω

1

, ω

2

g \ A, f ω

3

g \ A, f ω

4

g \ A g

= σ ff ω

1

, ω

2

g , f ω

3

g , ?g = f A, f ω

1

, ω

2

g , f ω

3

g , ?g . As a consequence, the measurable space on which

P ( jf X

2

= 1 g ) is built, is

( f ω

1

, ω

2

, ω

3

g , σ ff ω

1

, ω

2

g , f ω

3

gg ) .

(22)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability I

Independence

Let’s go back to the de…nition of independence between two random variables.

Theorem

The random variables X and Y are both built on the probability space ( , F , P ) , Card ( ) < . Let P

X

= f A

1

, ..., A

m

g and P

Y

= f B

1

, ..., B

n

g , be two …nite partitions that respectively generate the sigma-algebras σ ( X ) and σ ( Y ) . If

8 A 2 P

X

such that P ( A ) > 0, P ( B j A ) = P ( B ) , 8 B 2 P

Y

or again, if

8 B 2 P

Y

such that P ( B ) > 0, P ( A j B ) = P ( A ) , 8 A 2 P

X

,

then X and Y are independent.

(23)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability II

Independence

Proof of the theorem. Let’s assume that

8 A 2 P

X

such that P ( A ) > 0, P ( B j A ) = P ( B ) , 8 B 2 P

Y

. We want to show that 8 A 2 P

X

such that P ( A ) > 0 and 8 B 2 P

Y

,

P ( B \ A ) = P ( A ) P ( B ) . But 8 A 2 P

X

such that P ( A ) > 0 and 8 B 2 P

Y

,

P ( B ) = P ( B j A ) = P ( B \ A )

P ( A ) ) P ( B \ A ) = P ( A ) P ( B ) et 8 A 2 P

X

such that P ( A ) = 0 and 8 B 2 P

Y

,

0 P ( B \ A ) P ( A ) = 0 ) P ( B \ A ) = 0 = P ( A ) P ( B ) .

(24)

Independence Conditional probability

Sigma-algebra De…nition Interpretation Example Independence

Expectation Moments Conditional expectation

Conditional probability III

Independence

If we use

8 B 2 P

Y

such that P ( B ) > 0, P ( A j B ) = P ( A ) , 8 A 2 P

X

,

as a premise, we can follow a similar reasoning and obtain an

identical result.

(25)

Independence Conditional probability Expectation

De…nition Properties Example Remarks

Moments Conditional expectation

Expectation

De…nition

De…nition

Let X be a random variable built on the probability space ( Ω , F , P ) such that Card ( Ω ) < ∞ . The expectation of X , denoted E

P

[ X ] is

E

P

[ X ] = ∑

ω2Ω

X ( ω ) P ( ω ) . If the random variable X takes n di¤erent values, say x

1

< ... < x

n

then

E

P

[ X ] =

n i=1

x

i

P f ω 2 Ω : X ( ω ) = x

i

g =

n i=1

x

i

f

X

( x

i

)

where f

X

is the probability mass function of X .

(26)

Independence Conditional probability Expectation

De…nition Properties Example Remarks

Moments Conditional expectation

Expectation

Properties

Let X and Y be two random variables built on the probability space ( Ω , F , P ) , Card ( Ω ) < . If a and b represent real numbers, then

E1 E

P

[ aX + bY ] = aE

P

[ X ] + bE

P

[ Y ] ;

E2 If 8 ω 2 , X ( ω ) Y ( ω ) then E

P

[ X ] E

P

[ Y ] ; E3 In general, E

P

[ XY ] 6 = E

P

[ X ] E

P

[ Y ] ;

E4 If X and Y are independent, then

E

P

[ XY ] = E

P

[ X ] E

P

[ Y ] .

(27)

Independence Conditional probability Expectation

De…nition Properties Example Remarks

Moments Conditional expectation

Expectation I

Example

Example. Let’s take the random variable W again, as well as two probability measures P and Q :

ω W

(

ω

)

P

(

ω

)

Q

(

ω

)

1 5 16 124

2 5 16 121

3 5 16 121

4 5 16 121

5 0 16 121

6 10 16 124

x P

f

W

=

x

g

Q

f

W

=

x

g

0 16 121

5 46 127

10 16 124

(28)

Independence Conditional probability Expectation

De…nition Properties Example Remarks

Moments Conditional expectation

Expectation II

Example

EP

[

W

] = ∑

ω2Ω

W

(

ω

)

P

(

ω

)

=

5 1

6

+

5 1 6

+

5 1

6

+

5 1 6

+

0 1

6

+

10 1 6

=

30

6

=

5 EP

[

W

] =

n i=1

wifW

(

wi

) =

0 1 6

+

5 4

6

+

10 1 6

=

30

6

=

5 EQ

[

W

] = ∑

ω2Ω

W

(

ω

)

Q

(

ω

)

=

5 4

12

+

5 1

12

+

5 1

12

+

5 1

12

+

0 1

12

+

10 4 12

=

75

12

=

6,25 EQ

[

W

] =

n i=1

wifW

(

wi

) =

0 1

12

+

5 7

12

+

10 4 12

=

75

12

=

6,25

(29)

Independence Conditional probability Expectation

De…nition Properties Example Remarks

Moments Conditional expectation

Expectation

Remarks

The expectation of a random variable is a real number. It is not a random quantity.

From the previous example, we can notice that the

expectation of a random variable depends on the

probability measure that is being used.

(30)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Moments of a random variable

De…nition

De…nition

Let X be a random variable built on the probability space ( Ω , F , P ) such that Card ( Ω ) < ∞ . The kth moment of X , denoted E

P

X

k

is the expectation of the variable X

k

:

E

P

h X

k

i

= ∑

ω2Ω

X

k

( ω ) P ( ω ) =

n i=1

x

ik

f

X

( x

i

)

where x

1

< ... < x

n

are the values taken by X and f

X

is its

probability mass function.

(31)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Variance I

De…nition

De…nition

Let X be a random variable built on the probability space ( , F , P ) such that Card ( ) < . The variance of X , denoted Var

P

[ X ] is the expectation of the random variable

X E

P

[ X ]

2

:

Var

P

[ X ] = ∑

ω2Ω

X ( ω ) E

P

[ X ]

2

P ( ω ) (1)

=

n i=1

x

i

E

P

[ X ]

2

f

X

( x

i

) (2)

where x

1

< ... < x

n

are the values taken by X and f

X

is its

probability mass function.

(32)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Variance II

De…nition

Remarks.

The variance is a measure of dispersion of the values x

1

, ..., x

n

taken by X around the expectation E

P

[ X ] . The greater the variance, the more dispersed the values.

Similar to the expectation and the moments, the variance is a real number.

Moreover, whatever the random variable, the variance is never negative.

The standard deviation, commonly used in statistics, is the

square root of the variance.

(33)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Moments of a random variable

Variance properties

Exercise. Show that, if a and b are real numbers, V1 Var

P

[ X ] 0;

V2 Var

P

[ X ] = E

P

X

2

E

P

[ X ]

2

; V3 8 a 2 R , Var

P

[ aX + b ] = a

2

Var

P

[ X ] ; V4 If X and Y are independent, then

Var

P

[ X + Y ] = Var

P

[ X ] + Var

P

[ Y ]

(34)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Covariance I

De…nition

De…nition

Let X and Y be two random variables built on the probability space ( Ω , F , P ) such that Card ( Ω ) < ∞ . The covariance of X and Y , denoted Cov

P

[ X , Y ] is the expectation of the random variable X E

P

[ X ] Y E

P

[ Y ] :

Cov

P

[ X , Y ] = ∑

ω2Ω

X ( ω ) E

P

[ X ] Y ( ω ) E

P

[ Y ] P ( ω )

(35)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Covariance II

De…nition

Interpretation.

If the covariance is positive, it is because the sum

ω2Ω

X ( ω ) E

P

[ X ] Y ( ω ) E

P

[ Y ] P ( ω ) , is dominated by the ω that make the term

X ( ω ) E

P

[ X ] Y ( ω ) E

P

[ Y ] positive, which

means that the random variables X and Y tend to be

either greater or smaller than their respective expectations

in the same states of the world ω.

(36)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Covariance III

De…nition

If the covariance is negative, the sum has to be dominated by the ω which make the term

X ( ω ) E

P

[ X ] Y ( ω ) E

P

[ Y ] negative, which

means that, when one of the random variables X and Y is

greater than its expectation, the other one tends to be

smaller than its own expectation.

(37)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Covariance

Properties

Exercise. Show that

C1 Cov

P

[ X , Y ] = E

P

[ XY ] E

P

[ X ] E

P

[ Y ] ;

C2 If X and Y are independent, then Cov

P

[ X , Y ] = 0;

C3 8 a, b 2 R ,

Cov

P

[ aX

1

+ bX

2

; Y ] = aCov

P

[ X

1

; Y ] + bCov

P

[ X

2

; Y ] ;

C4 Var

P

[ X + Y ] = Var

P

[ X ] + Var

P

[ Y ] + 2Cov

P

[ X , Y ] .

(38)

Independence Conditional probability Expectation Moments

De…nition Variance

Properties Covariance Properties Conditional expectation

Other moments

De…ntion

In addition to the variance, two other centered moments are commonly used in modelling: skewness and kurtosis

coe¢ cients. A description thereof can be found in the appendix

to this document.

(39)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation

De…nition

De…nition

The random variable X is built on the probability space ( , F , P ) , Card ( ) < . Let G F , a sigma-algebra generated by the …nite partition P = f A

1

, ..., A

n

g satisfying 8 i 2 f 1, ..., n g , P ( A

i

) > 0. The conditional expectation of X given G , denoted E

P

[ X jG ] is

E

P

[ X jG ] ( ω ) =

n i=1

I

Ai

( ω ) P ( A

i

) ∑

ω 2Ai

X ( ω ) P ( ω ) where I

Ai

: Ω ! f 0, 1 g is the indicator function

I

Ai

( ω ) = 1 si ω 2 A

i

0 si ω 2 / A

i

.

(40)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation I

Example

Example . Let’s go back to the stochastic process representing the share price of a security, to which we will add a probability measure on the measurable space ( Ω , F ) .

ω X

0

( ω ) X

1

( ω ) X

2

( ω ) X

3

( ω ) P ( ω ) ω

1

1

12

1

12 18

ω

2

1

12

1

12 28

ω

3

1 2 1 1

38

ω

4

1 2 2 2

28

(41)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation II

Example

We had determined that F

1

= σ ff ω

1

, ω

2

g , f ω

3

, ω

4

gg . Then

EP

[

X3

jF

1

] (

ω

)

=

2 i=1

IAi

(

ω

)

P

(

Ai

) ∑

ω 2Ai

X3

(

ω

)

P

(

ω

)

=

IA1

(

ω

)

P

(

A1

) ∑

ω2A1

X3

(

ω

)

P

(

ω

) +

IA2

(

ω

)

P

(

A2

) ∑

ω2A2

X3

(

ω

)

P

(

ω

)

=

IA1

(

ω

)

3 8

1 2

1 8

+

1

2 2

8

+

IA2

(

ω

)

5 8

1 3 8

+

2 2

8

=

1

2IA1

(

ω

) +

7

5IA2

(

ω

)

.

(42)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation III

Example

Thus,

If ω 2 f ω

1

, ω

2

g then E

P

[ X

3

jF

1

] ( ω ) = 1

2 I

A1

( ω ) + 7

5 I

A2

( ω ) = 1 2 and if ω 2 f ω

3

, ω

4

g then

E

P

[ X

3

jF

1

] ( ω ) = 1

2 I

A1

( ω ) + 7

5 I

A2

( ω ) = 7 5 . Interpretation. At time t = 1, we will be able to

determine whether the state of the world is an element of

f ω

1

, ω

2

g or an element of f ω

3

, ω

4

g . If ω 2 f ω

1

, ω

2

g

then the expected value of the random variable X

3

is

12

.

By contrast, if ω 2 f ω

3

, ω

4

g , then the expected value of

X

3

is

75

.

(43)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation I

Remarks

The conditional expectation is expressed in terms of the conditional probabilities given each of the elements in the partition which generates the sigma-algebra, since

E

P

[ X jG ] ( ω )

=

n i=1

I

Ai

( ω ) P ( A

i

) ∑

ω 2Ai

X ( ω ) P ( ω )

=

n i=1

I

Ai

( ω ) ∑

ω 2Ai

X ( ω ) P ( ω \ A

i

) P ( A

i

)

=

n i=1

I

Ai

( ω ) ∑

ω 2Ai

X ( ω ) P ( ω j A

i

)

(44)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation II

Remarks

Important remark. Unlike the expectation, the

conditional expectation is not a real number, but a random

variable. Actually, since it is constant on the atoms which

generate G , it is a G measurable random variable.

(45)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation I

Properties

Let X and Y be two random variables in the probability space ( Ω , F , P ) .

The sigma-algebras G , G

1

and G

2

are respectively

generated by the …nite partitions P = f A

1

, ..., A

n

g and

P

1

= f B

1

, ..., B

m

g and P

2

= f C

1

, ..., C

n

g .

(46)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation II

Properties

EC1 If X is G measurable, then E

P

[ X jG ] = X . EC2 If G

1

G

2

are sigma-algebras, then

E

P

E

P

[ X jG

1

] jG

2

= E

P

[ X jG

1

] . EC3 If G

1

G

2

are sigma-algebras, then

E

P

E

P

[ X jG

2

] jG

1

= E

P

[ X jG

1

] . EC4 E

P

[ X jf? , g ] = E

P

[ X ] .

EC5 E

P

E

P

[ X jG ] = E

P

[ X ] .

EC6 If Y is G measurable, then E

P

[ XY jG ] = Y E

P

[ X jG ] . EC7 If X and Y are independent, then

E

P

[ X j σ ( Y ) ] = E

P

[ X ] .

EC8 8 a, b 2 R , E

P

[ aX + bY jG ] = aE

P

[ X jG ] + bE

P

[ Y jG ] .

(47)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation I

Proof of EC1

To be shown. If X is G measurable, then E

P

[ X jG ] = X .

Since X is G measurable, then it is constant on the atoms of G . So, let’s set

x

i

= X ( ω ) 8 ω 2 A

i

, 8 i 2 f 1, ..., n g .

Let’s consider any ω 2 A

i

.

(48)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation II

Proof of EC1

E

P

[ X jG ] ( ω ) =

n j=1

I

Aj

( ω ) P ( A

j

) ∑

ω 2Aj

X ( ω ) P ( ω )

= 1

P ( A

i

) ∑

ω 2Ai

X ( ω ) P ( ω )

= 1

P ( A

i

) ∑

ω 2Ai

x

i

P ( ω )

= x

i

P ( A

i

) ∑

ω 2Ai

P ( ω )

= x

i

P ( A

i

) P ( A

i

)

= x

i

= X ( ω ) .

(49)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation III

Proof of EC1

Since ω was arbitrarily chosen, we have that

8 ω 2 Ω , E

P

[ X jG ] ( ω ) = X ( ω ) .

(50)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation

Proof of EC2

To be shown, If G

1

G

2

F are sigma-algebras,then E

P

E

P

[ X jG

1

] jG

2

= E

P

[ X jG

1

] .

Since G

1

G

2

, then any G

1

measurable function is also G

2

measurable. Moreover, we know that E

P

[ X jG

1

] is G

1

measurable, therefore E

P

[ X jG

1

] is also G

2

measurable.

By using ( EC 1 ) , E

P

h

E

P

[ X jG

1

] jG

2

i = E

P

[ X jG

1

] .

(51)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation I

Proof of EC3

To be shown. If G

1

G

2

F are sigma-algebras, then E

P

E

P

[ X jG

2

] jG

1

= E

P

[ X jG

1

] .

Let B

k

be a atom of G

1

. Since B

k

2 G

1

G

2

then B

k

2 G

2

and, as a consequence, B

k

can be represented as the union of some atoms of G

2

, i.e. there exist C

k1

, ..., C

kq

2 P

2

such that B

k

= S

qi=1

C

ki

. Moreover, notice that, E

P

[ X jG

2

] being G

2

measurable, it is constant on the atoms of G

2

. Let’s set 8 k 2 f 1, ..., n g , 8 ω 2 C

k

,

x

k

E

P

[ X jG

2

] ( ω )

=

n j=1

I

Cj

( ω ) P ( C

j

) ∑

ω 2Cj

X ( ω ) P ( ω )

= 1

P ( C

k

) ∑

ω 2Ck

X ( ω ) P ( ω ) .

(52)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation II

Proof of EC3

Let’s consider any ω 2 C

k0

B

k

. E

P

h

E

P

[ X jG

2

] jG

1

i ( ω )

=

m j=1

I

Bj

( ω ) P ( B

j

) ∑

ω 2Bj

E

P

[ X jG

2

] ( ω ) P ( ω )

= 1

P ( B

k

) ∑

ω 2Bk

E

P

[ X jG

2

] ( ω ) P ( ω )

= 1

P ( B

k

)

q i=1

ω 2Cki

E

P

[ X jG

2

] ( ω )

| {z }

xki

P ( ω )

= 1

P ( B

k

)

q i=1

x

ki

ω 2Cki

P ( ω )

= 1

P ( B

k

)

q i=1

x

ki

P ( C

ki

)

(53)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation III

Proof of EC3

= 1

P ( B

k

)

q i=1

0

@ 1

P ( C

ki

) ∑

ω 2Cki

X ( ω ) P ( ω ) 1

A P ( C

ki

)

= 1

P ( B

k

)

q i=1

ω 2Cki

X ( ω ) P ( ω )

= 1

P ( B

k

) ∑

ω 2Bk

X ( ω ) P ( ω )

=

m j=1

I

Bj

( ω ) P ( B

j

) ∑

ω 2Bj

X ( ω ) P ( ω )

= E

P

[ X jG

1

] ( ω ) .

Thus, 8 ω 2 Ω , E

P

E

P

[ X jG

2

] jG

1

( ω ) = E

P

[ X jG

1

] ( ω ) .

(54)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation

Proof of EC4

To be shown. E

P

[ X jf? , Ω g ] = E

P

[ X ] .

Since f? , g is a sigma-algebra generated by Ω , by applying the de…nition of the conditional expectation of X with respect to such a sigma-algebra, we obtain 8 ω 2 Ω ,

E

P

[ X jf? , Ω g ] ( ω ) =

1 j=1

I

( ω ) P ( Ω ) ∑

ω 2Ω

X ( ω ) P ( ω )

= ∑

ω 2Ω

X ( ω ) P ( ω )

= E

P

[ X ] .

(55)

Independence Conditional probability Expectation Moments Conditional expectation De…nition Example Properties

Conditional expectation

Proof of EC5

To be shown. E

P

E

P

[ X jG ] = E

P

[ X ] .

Just apply Property ( EC 3 ) with G

2

= G and G

1

= f? , Ω g .

Références

Documents relatifs

Sample space Random variables Probability measures Distributions (laws) Sigma- algebras Probability measures (continued) References Appendices..

We already established that it is not possible to build a prob- ability measure on the sample space such that both random variables X and W have a uniform distribution?. Is it

Exercice 2.2. You toss a coin four times. On the …rst toss, you win 10 dollars if the result is “tails” and lose 10 dollars if the result is “heads”. On the subsequent tosses,

We have seen that, if the model contains no arbitrage opportunities, then there exists a probability measure Q such that the contingent claim price at time 0 is the expectation of

normal law Properties Other properties Construction of the Brownian motion An alternative

able species. To this aim the idea of an event is defined and the basic properties of the theory are derived. Two kinds of stochastic variables are introduced. For

In particular we show that risk premia decompose exactly into the sum of two terms, the correlation of returns with a market density process and a component arising from the lack

When the vector field Z is holomorphic the real vector fields X and Y (cf. This follows from a straightforward computation in local coordinates.. Restriction to the