• No results found

Measure Theory and Probability

N/A
N/A
Protected

Academic year: 2022

Share "Measure Theory and Probability"

Copied!
39
0
0

Loading.... (view fulltext now)

Full text

(1)

Measure Theory and Probability

Module VI: Continuity Theorem

Dr. Arindom Chakraborty

October 31, 2015

(2)

Objectives

I Concept of continuity for a sequence of sets

I Continuity theorem of measure and Probability

(3)

Objectives

I Concept of continuity for a sequence of sets

I Continuity theorem of measure and Probability

(4)

Statement

I Let λbe an additive set function defined on (Ω,F). Then

1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A) is continuous from above)

λis continuousλis continuous from above and below 3. λis continuous from belowλis σ−additive

4. λis finite and continuous atφλisσ−additive

(5)

Statement

I Let λbe an additive set function defined on (Ω,F). Then 1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A) is continuous from above)

λis continuousλis continuous from above and below 3. λis continuous from belowλis σ−additive

4. λis finite and continuous atφλisσ−additive

(6)

Statement

I Let λbe an additive set function defined on (Ω,F). Then 1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A)

is continuous from above)

λis continuousλis continuous from above and below 3. λis continuous from belowλis σ−additive

4. λis finite and continuous atφλisσ−additive

(7)

Statement

I Let λbe an additive set function defined on (Ω,F). Then 1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A) is continuous from above)

λis continuousλis continuous from above and below 3. λis continuous from belowλis σ−additive

4. λis finite and continuous atφλisσ−additive

(8)

Statement

I Let λbe an additive set function defined on (Ω,F). Then 1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A) is continuous from above)

λis continuousλis continuous from above and below

3. λis continuous from belowλis σ−additive 4. λis finite and continuous atφλisσ−additive

(9)

Statement

I Let λbe an additive set function defined on (Ω,F). Then 1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A) is continuous from above)

λis continuousλis continuous from above and below 3. λis continuous from belowλis σ−additive

4. λis finite and continuous atφλisσ−additive

(10)

Statement

I Let λbe an additive set function defined on (Ω,F). Then 1. λisσ−additive =forAn∈ F ↑n,

limλ(An) =λ(limAn) =λ(∪n=1An) =λ( ¯A) is continuous from below)

2. λisσ−additive =forAn∈ F ↓n, λ(A1)< limλ(An) =λ(limAn) =λ(∩n=1An) =λ(A) is continuous from above)

λis continuousλis continuous from above and below 3. λis continuous from belowλis σ−additive

4. λis finite and continuous atφλisσ−additive

(11)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞ n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(12)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞ n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(13)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(14)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(15)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(16)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(17)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An]

λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(18)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An)

λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(19)

Proof

I 1. LetAnn

LetBi = AC1AC2...ACi−1Ai, then i=1Ai=i=1Bi

λ(∪i=1Ai) = λ(∪i=1Bi) =λ(∪i=1AC1AC2...ACi−1Ai)

=

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞

n

X

i=1

λ(AC1AC2...ACi−1Ai)

= lim

n→∞λ(∪ni=1AC1AC2...ACi−1Ai)[λisσadditive]

= lim

n→∞λ(∪ni=1Ai) = lim

n→∞λ(An) 2. LetAnndefineBn=A1An,n

λ(∪i=1Bi) = lim

n→∞λ(Bn)[ by previous result ]

= lim

n→∞λ(A1An)[A1= (A1An)An] λ(∪i=1(A1Ai)) = λ(A1) lim

n→∞λ(An) λ(A1)λ(∩i=1Ai) = λ(A1) lim

n→∞λ(An)

(20)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn=∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞ n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(21)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn)

= lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞ n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(22)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞ n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(23)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞

n

X

i=1

λ(Ai)[λis additive ]

=

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(24)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞

n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0

LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(25)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞

n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(26)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞

n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

(27)

Proof

I 3. Let A1,A2, ...,An∈ F be a sequence of pairwise disjoint sets. Define

Bn=∪ni=1Ai ↑n. Then

limn→∞Bn=∪n=1Bn =∪i=1Ai

λ(∪i=1Ai) = λ( lim

n→∞Bn) = lim

n→∞λ(Bn)[λcontinuous from below ]

= lim

n→∞λ(∪ni=1Bi) = lim

n→∞

n

X

i=1

λ(Ai)[λis additive ] =

X

i=1

λ(Ai)

I 4. Let Bn↓φ. Then

limn→∞λ(Bn) =λ(limn→∞Bn) =λ(φ) = 0 LetAi ∩Aj =φ. DefineBn= ¯A−(∪ni=1Ai)↓n

n→∞lim λ(Bn) =λ( lim

n→∞Bn) = 0

⇔ lim

n→∞λ( ¯A−(∪ni=1Ai)) = λ( lim

n→∞( ¯A−(∪ni=1Ai)))

⇔ lim

n→∞[λ( ¯A)−λ(∪ni=1Ai)] = 0⇔λ( ¯A) = lim

n→∞ n

X

i=1

λ(Ai)

References

Related documents

1 Two very efficient algorithms for training structured models that avoids the problem of exponential output space. 2

We are now ready to see how the forward and backward probabilities can help compute the transition probability a i j and observation probability b i (o t ) from an ob-

where jEj is the size (cardinality) of the set E i.e. number of outcomes in favour of event E and jj is the total number of outcomes possible. Of course, the inherent assumption is

I to find limit of a sequence of sets, if exists, using some examples. I to define Indicator function

Trajectory space (σ- algebra generated by basic cylinders of equiv class of trajectories).. Probability measure for the

Assumed a Geometric Graph model of a Wireless Sensor Network HD is not a good measure of ED for arbitrary node placement Established high probability bounds on the ED, given the HD

Lkkjka'k % eSllZ bafM;u esVy Qsjks ,sykW;] mM+hlk dh ØksekbV [kqyh [knku esa Mai &lt;ykuksa ds &lt;yku fMT+kkbu gsrq Hkw&amp;rduhdh v/;;u fd;k x;kA Mai dh lkefxz;ksa ds

A study on residual Kull- back Leibler divergence measure, Renyi’s divergence measure and Kerridge’s inaccuracy measure for conditionally specified and conditional survival rvs