Chapter 2. Axioms of Probability
欢迎大家来我的github下载源码呀,https://github.com/Berry-Wen/statistics-note-system
σ
−
a
l
g
e
b
r
a
\sigma-algebra
σ−algebra
-
Let Ω \Omega Ω be abstract space, that is with no special structure.
-
Let 2 Ω 2^{\Omega} 2Ω denotes all subsets of Ω \Omega Ω , including the empty set denoted by ∅ \emptyset ∅.
With A \mathcal{A} A being a subset of 2 Ω 2^{\Omega} 2Ω , we consider the following properties:
-
∅ ∈ A a n d Ω ∈ A \emptyset \in \mathcal{A} \quad and \quad \Omega \in \mathcal{A} ∅∈AandΩ∈A
-
If A ∈ A A \in \mathcal{A} A∈A then A C ∈ A A^C \in \mathcal{A} AC∈A , where A C A^C AC denotes the complement of A A A
-
A \mathcal{A} A is closed under finite unions and finite intersections:
- that is, if A 1 , . . . , A n A_1,...,A_n A1,...,An are all in A \mathcal{A} A , then ∪ i = 1 n A i \cup_{i=1}^{n}A_i ∪i=1nAi and ∪ i = 1 n A i \cup_{i=1}^nA_i ∪i=1nAi are in A \mathcal{A} A as well
- A 1 , . . . , A n ∈ A ⇒ ∪ i = 1 n A i a n d ∩ i = 1 n A i ∈ A A_1,...,A_n \in \mathcal{A} \qquad \Rightarrow \qquad \cup_{i=1}^n A_i \quad and \quad \cap_{i=1}^n A_i \in \mathcal{A} A1,...,An∈A⇒∪i=1nAiand∩i=1nAi∈A
-
A \mathcal{A} A is closed under countable unions and intersections:
- that is, if A 1 , A 2 . . . A_1,A_2... A1,A2... is countable sequence of events in A \mathcal{A} A , then ∪ i = 1 n A i \cup_{i=1}^n A_i ∪i=1nAi and ∩ i = 1 n A i \cap_{i=1}^n A_i ∩i=1nAi are both also in A \mathcal{A} A
Definition 2.1
A \mathcal{A} A is an algebra if it satisfies (1),(2) and (3) above
A \mathcal{A} A is an σ − a l g e b e r a \sigma-algebera σ−algebera (or a σ − f i e l d \sigma-field σ−field), if is satisfies (1),(2) and (4) above
- Under (2), (1) can replaced by either (1’): ∅ ∈ A \emptyset \in \mathcal{A} ∅∈A or by (1’’): Ω ∈ A \Omega \in \mathcal{A} Ω∈A
- (1)+(4) ⇒ \Rightarrow ⇒ (3) hence any σ − a l g e b r a \sigma-algebra σ−algebra is an a l g e b r a algebra algebra (but there are a l g e b r a algebra algebra that are not σ − a l g e b r a \sigma-algebra σ−algebra)
Exercise 2.17 Suppose that Ω \Omega Ω is an infinite set (countable or not), and let A \mathcal{A} A be the family of all subsets which are either finite or have a finite complement. Show that A \mathcal{A} A is an a l g e b r a algebra algebra , but not a σ − a l g e b r a \sigma-algebra σ−algebra
Solution:
Under the definition, A \mathcal{A} A is the family of subsets which are either finite or have a finite complement.
-
∅ \emptyset ∅ is a finite subset ⇒ ∅ ∈ A \Rightarrow \quad \emptyset \in \mathcal{A} ⇒∅∈A
-
if A ∈ A ⇒ A A \in \mathcal{A} \Rightarrow \quad A A∈A⇒A is finite or A C A^C AC is finite
so A C ∈ A A^C \in \mathcal{A} AC∈A
-
if A i ∈ A , i = 1 , 2 , . . . , n ⇒ A i A_i \in \mathcal{A}, i=1,2,...,n \quad \Rightarrow \quad A_i Ai∈A,i=1,2,...,n⇒Ai is finite or A C A^C AC is finite
- if ∃ i 0 \quad \exist i_0 ∃i0 so that A i 0 A_{i_{0}} Ai0 is finite
∩ i = 1 n A i ⊂ A i 0 is finite ⇓ so that ∩ i = 1 n A i is finite \cap_{i=1}^n A_i \subset A_{i_0} \quad \text{is finite} \\ \Downarrow \\ \text{so that} \quad \cap_{i=1}^n A_i \quad \text{is finite} ∩i=1nAi⊂Ai0is finite⇓so that∩i=1nAiis finite
-
∀ i = ( 1 , 2 , . . . , n ) \forall i=(1,2,...,n) ∀i=(1,2,...,n) , A i A_i Ai is infinite
⇒ A i C \Rightarrow A_i^C ⇒AiC is finite
( ∩ i = 1 n A i ) C = ∪ i = 1 n A i C finite ⇓ s o ( ∩ i = 1 n A i ) C ∈ A \left( \cap_{i=1}^n A_i \right)^C = \cup_{i=1}^n A_i^C \quad \text{finite} \\ \Downarrow \\ so \quad \left( \cap_{i=1}^n A_i \right)^C \in \mathcal{A} (∩i=1nAi)C=∪i=1nAiCfinite⇓so(∩i=1nAi)C∈A
-
Ω \Omega Ω is infinite , there exist ω ∈ Ω i = 1 , 2 , ⋯ \omega \in \Omega \quad i=1,2,\cdots ω∈Ωi=1,2,⋯
记 A i = { ω i } i = 1 , 2 , ⋯ ⇒ A i ∈ A A_i= \left\{ \omega_i \right\} \quad i=1,2,\cdots \quad \Rightarrow \quad A_i \in \mathcal{}A Ai={ωi}i=1,2,⋯⇒Ai∈A
Consider ∪ m = 1 ∞ = { ω 2 m , ω 2 m + 2 , ⋯ } \cup_{m=1}^{ \infty}= \left\{ \omega_{2m},\omega_{2m+2},\cdots \right\} ∪m=1∞={ω2m,ω2m+2,⋯} is infinite
Moreover , for each fixed m = 1 , 2 , ⋯ m=1,2,\cdots m=1,2,⋯ , we have ω 2 m − 1 ∈ A 2 m C \omega_{2m-1} \in A_{2m}^{C} ω2m−1∈A2mC for every m = 1 , 2 , ⋯ m=1,2,\cdots m=1,2,⋯
So A 2 m − 1 ⊂ A 2 m C A_{2m-1} \subset A_{2m}^C A2m−1⊂A2mC for every m = 1 , 2 , ⋯ m=1,2,\cdots m=1,2,⋯
∪ m = 1 ∞ A 2 m − 1 ⊂ A 2 m C \cup_{m=1}^{ \infty}A_{2m-1} \subset A_{2m}^C ∪m=1∞A2m−1⊂A2mC
∴ infinite ∪ m = 1 ∞ A 2 m − 1 ⊂ ∩ m = 1 ∞ A 2 m C = ( ∪ m = 1 ∞ A 2 m C ) C \therefore \quad \text{infinite} \quad\cup_{m=1}^{ \infty}A_{2m-1} \subset \cap_{m=1}^{ \infty}A_{2m}^C= \left( \cup_{m=1}^{ \infty}A_{2m}^C \right)^C \quad ∴infinite∪m=1∞A2m−1⊂∩m=1∞A2mC=(∪m=1∞A2mC)C is infinite
∴ ∪ m = 1 ∞ A 2 m \therefore \quad \cup_{m=1}^{ \infty}A_{2m} \quad ∴∪m=1∞A2m if infinite
∴ ( ∪ m = 1 ∞ A 2 m ) C \therefore \quad \left(\cup_{m=1}^{ \infty}A_{2m} \right)^C\quad ∴(∪m=1∞A2m)C is infinite
Definition 2.2
If C ⊂ 2 Ω \mathcal{C} \subset 2^{\Omega} C⊂2Ω , The σ − a l g e b r a \sigma-algebra σ−algebra generated by C \mathcal{C} C , and written as σ ( C ) \sigma(\mathcal{C}) σ(C).
- 2 Ω 2^{\Omega} 2Ω is a σ − a l g e b r a \sigma-algebra σ−algebra
- Exercise 2.2 Let ( G α ) α ∈ A (\mathcal{G}_{\alpha})_{\alpha \in A} (Gα)α∈A be an arbitrary family of σ − a l g e b r a s \sigma-algebras σ−algebras defined on abstract space Ω \Omega Ω. Show that H = ∩ α ∈ A G α \mathcal{H}= \cap_{\alpha \in A}\mathcal{G}_{\alpha} H=∩α∈AGα is also a σ − a l g e b r a \sigma-algebra σ−algebra.
Example
- A = { ∅ , Ω } \mathcal{A}= \left\{ \emptyset,\Omega \right\} A={∅,Ω} (the trivial σ − a l g e b r a \sigma-algebra σ−algebra)
- A A A is a subset, then σ ( A ) = { ∅ , A , A C , Ω } \sigma(A)= \left\{ \emptyset ,A ,A^C,\Omega \right\} σ(A)={∅,A,AC,Ω}
- if Ω = R \Omega = \mathbb{R} Ω=R (the real numbers) (or more generally if Ω \Omega Ω is a space with a topology, a case we treat in Chapter 8), the Borel σ − a l g e b r a \sigma-algebra σ−algebra is the σ − a l g e b r a \sigma-algebra σ−algebra generated by the open sets (or by the closed sets, which is equivalent)
trivial 平凡的
Borel σ − a l g e b r a \sigma -algebra σ−algebra
博雷尔集—由所有开集生成的 σ \sigma σ 代数,或包含所有闭集生成的 σ \sigma σ 代数
Wolfram Mathworld: A sigma-algebra which is related to the topology of a set. The Borel sigma-algebra is defined to be the sigma-algebra generated by the open sets (or equivalently, by the closed sets).
Wikipedia:
-
The Borel algebra on X is the smallest σ-algebra containing all open sets (or, equivalently, all closed sets).
-
Borel sets are important in measure theory, since any measure defined on the open sets of a space, or on the closed sets of a space, must also be defined on all Borel sets of that space. Any measure defined on the Borel sets is called a Borel measure. Borel sets and the associated Borel hierarchy also play a fundamental role in descriptive set theory.
作业:(补补补)
Exercise 2.2 Let ( G α ) α ∈ A (\mathcal{G}_{\alpha})_{\alpha \in A} (Gα)α∈A be an arbitrary family of σ − a l g e b r a s \sigma-algebras σ−algebras defined on abstract space Ω \Omega Ω. Show that H = ∩ α ∈ A G α \mathcal{H}= \cap_{\alpha \in A}\mathcal{G}_{\alpha} H=∩α∈AGα is also a σ − a l g e b r a \sigma-algebra σ−algebra.
-
Ω \Omega Ω 是不是 H \mathcal{H} H
-
∀ B ∈ H = ∩ α ∈ Λ G α \forall B \in \mathcal{H}= \cap_{\alpha \in \Lambda}\mathcal{G}_{\alpha} ∀B∈H=∩α∈ΛGα
∀ α ∈ A B ∈ G α is sigma algebra \forall \alpha \in A \quad B \in \mathcal{G}_{\alpha} \quad \text{is sigma algebra} ∀α∈AB∈Gαis sigma algebra
so B c ∈ G α \text{so} \quad B^c \in \mathcal{G}_{\alpha} soBc∈Gα
B c ∈ ∩ α ∈ Λ G α = H B^c \in \cap_{\alpha \in \Lambda} \mathcal{G}_{\alpha} = \mathcal{H} Bc∈∩α∈ΛGα=H
- 2 Ω 2^{\Omega} 2Ω 样本空间的最大 σ \sigma σ 代数
- C \mathcal{C} C 是集合类
- σ ( C ) \sigma(\mathcal{C}) σ(C) 是最小的 σ \sigma σ 代数
相关概念:
- 内点:如果存在 P 0 P_0 P0 的某一邻域 U ( P 0 ) U(P_0) U(P0) ,使得 U ( P 0 ) ⊂ E U(P_0) \subset E U(P0)⊂E ,则称 P 0 P_0 P0 为 E E E 的内点
- 外点:如果 P 0 P_0 P0是 E c E^c Ec 的内点,则称 P 0 P_0 P0为 E E E 的外点(这里的余集是针对全空间 R n R^n Rn 来作的
- 界点:如果 P 0 P_0 P0 既非 E E E 的内点又非 E E E 的外点,即: P 0 P_0 P0 的任一邻域内既有属于 E E E 的点,也有不属于 E E E 的点
- 聚点:设 E E E 是 R n R^n Rn 中一点集, P 0 P_0 P0 为 R n R^n Rn 中一定点,如果 P 0 P_0 P0 的任一邻域内都含有无穷多个属于 E E E 的点,则称 P 0 P_0 P0 为 E E E 的一个聚点
- 孤立点:如果 P 0 P_0 P0 属于 E E E 但不是 E E E 的聚点,则 P 0 P_0 P0 为 E E E 的孤立点
-
开核: E E E 的全体内点所组成的集合,记为 E ˚ \mathring{E} E˚
E ˚ = { x : ∃ U ( x ) ⊂ E } \mathring{E} = \left\{ x: \exists U(x) \subset E \right\} E˚={x:∃U(x)⊂E}
-
导集: E E E 的全体聚点所组成的集合,记为 E ′ E' E′
E ′ = { x : ∀ U ( x ) , U ( x ) ∩ E \ { x } ≠ ∅ } E' = \left\{ x: \forall U(x), U(x) \cap E \backslash \{x\} \neq \emptyset \right\} E′={x:∀U(x),U(x)∩E\{x}=∅}
-
边界: E E E 的全体界点所组成的集合,记为 ∂ E \partial E ∂E
∂ E = { x : ∀ U ( x ) , U ( x ) ∩ E ≠ ∅ , U ( x ) ∩ E c ≠ ∅ } \partial E = \left\{ x: \forall U(x),U(x) \cap E \neq \emptyset ,U(x) \cap E^c \neq \emptyset \right\} ∂E={x:∀U(x),U(x)∩E=∅,U(x)∩Ec=∅}
-
闭包: E ∪ E ′ E \cup E' E∪E′ 记为 E E E 的闭包,记为 E ‾ \overline{E} E
E ‾ = { x : ∀ U ( x ) , U ( x ) ∩ E ≠ ∅ } = E ∪ ∂ E = E ˚ ∪ ∂ E = E ′ ∪ { E 的孤立点 } \begin{aligned} \overline{E} &= \left\{ x: \forall U(x), U(x) \cap E \neq \emptyset \right\} \\ &= E \cup \partial E \\ &= \mathring{E} \cup \partial E \\ &= E' \cup \left\{ E \text{的孤立点} \right\} \\ \end{aligned} E={x:∀U(x),U(x)∩E=∅}=E∪∂E=E˚∪∂E=E′∪{E的孤立点}
-
开集:设 E ⊂ R n E \subset R^n E⊂Rn ,如果 E E E 的每一个点都是 E E E 的内点,则称 E E E 为开集
-
闭集:设 E ⊂ R n E \subset R^n E⊂Rn ,如果 E E E 的每一个聚点都属于 E E E ,则称 E E E 为闭集
开集与闭集的对偶性:
E 为开集,则 E c 为闭集 E \text{为开集,则} E^c \text{为闭集} E为开集,则Ec为闭集
Theorem 2.1
The Borel σ − a l g e b r a \sigma-algebra σ−algebra of R \mathbb{R} R is generated by intervals of the form ( − ∞ , a ] (- \infty,a] (−∞,a], where a ∈ Q a \in \mathbb{Q} a∈Q
( Q = r a t i o n a l s ) \quad (\mathbb{Q} = rationals) (Q=rationals)
Proof.
Let C \mathcal{C} C denote all open intervals. Since every open set in R \mathcal{R} R is the countable union of open intervals, we have σ ( C ) = \sigma(\mathcal{C})= σ(C)= the Borel σ − a l g e b r a \sigma-algebra σ−algebra of R \mathbb{R} R
Let D \mathcal{D} D denote all intervals of the form ( − ∞ , a ] (- \infty,a] (−∞,a], where a ∈ Q a \in \mathbb{Q} a∈Q.
Let ( a , b ) ∈ C (a,b) \in \mathcal{C} (a,b)∈C
Let ( a n ) n ≥ 1 (a_n)_{n \ge 1} (an)n≥1 be a sequence of rationals decreasing to a
Let ( b n ) n ≥ 1 (b_n)_{n \ge_1} (bn)n≥1 be a sequence of rationals increasing strictly to b.
Then
( a , b ) = ∪ n = 1 ∞ ( a n , b n ] = ∪ n = 1 ∞ ( ( − ∞ , b n ] ∩ ( − ∞ , a n ] c ) \begin{aligned} (a,b) &= \cup_{n=1}^{ \infty} (a_n, b_n] \\ &= \cup_{n=1}^{ \infty} \left( (- \infty,b_n] \cap (- \infty,a_n]^c \right) \\ \end{aligned} (a,b)=∪n=1∞(an,bn]=∪n=1∞((−∞,bn]∩(−∞,an]c)
Therefore , C ⊂ σ ( D ) \mathcal{C} \subset \sigma(\mathcal{D}) C⊂σ(D), where σ ( C ) ⊂ σ ( D ) \sigma(\mathcal{C}) \subset \sigma(\mathcal{D}) σ(C)⊂σ(D)
However, since each element of
D
\mathcal{D}
D is a closed set, it is also a Borel set, and therefore
D
\mathcal{D}
D is contained in the Borel sets
B
\mathcal{B}
B, Thus we have
B
=
σ
(
C
)
⊂
σ
(
D
)
⊂
B
\mathcal{B} = \sigma(\mathcal{C}) \subset \sigma(\mathcal{D}) \subset \mathcal{B}
B=σ(C)⊂σ(D)⊂B
and hence σ ( D ) = B \sigma(\mathcal{D}) = \mathcal{B} σ(D)=B
这个定理给出了更容易验证 B o r e l σ − a l g e b r a Borel \sigma-algebra Borelσ−algebra 的一种方法,给出了验证博雷尔集的一个充要条件
Definition 2.3
A probability measure defined on a σ − a l g e b r a A \sigma-algebra \quad \mathcal{A} σ−algebraA of Ω \Omega Ω is a function P : A → [ 0 , 1 ] P: \mathcal{A} \to [0,1] P:A→[0,1] that satisfies:
-
P ( Ω ) = 1 P(\Omega)=1 \quad P(Ω)=1 规范性
-
For every countable sequence ( A n ) n ≥ 1 (A_n)_{n\ge 1} (An)n≥1 of elements of A \mathcal{A} A, pairwise disjoint (that is, A n ∩ A m = ∅ A_n \cap A_m= \emptyset An∩Am=∅ Whenever n ≠ m n \neq m n=m), one has:
P ( ∪ n = 1 ∞ A n ) = ∑ n = 1 ∞ P ( A n ) P \left( \cup_{n=1}^{ \infty} A_n \right) = \sum_{n=1}^{\infty} P \left( A_n \right) P(∪n=1∞An)=n=1∑∞P(An)
Axiom 2 above is called countable additivity; the number P ( A ) P(A) P(A) is called the probability of the event A A A
- 有限可加 additivity
- 可列可加 countable additivity
概率是 A \mathcal{A} A 到 Ω \Omega Ω 上的一个映射,是任何一个样本空间的子集
Remark: In definition 2.3 one might imagine a more naive condition than 2, namely
A
,
B
∈
A
,
A
∩
B
=
∅
⇒
P
(
A
∪
B
)
=
P
(
A
)
+
P
(
B
)
A,B \in \mathcal{A} , A \cap B = \emptyset \quad \Rightarrow \quad P (A \cup B) = P(A) + P(B)
A,B∈A,A∩B=∅⇒P(A∪B)=P(A)+P(B)
This property is called additivity (or “finite additivity”) and, by an elementary induction, it implies that for every finite
A
1
,
⋯
,
A
m
A_1,\cdots,A_m
A1,⋯,Am of pairwise disjoint events
A
i
∈
A
A_i \in \mathcal{A}
Ai∈A we have
P
(
∪
n
=
1
m
A
N
)
=
∑
n
=
1
∞
P
(
A
n
)
P \left( \cup_{n=1}^{m} A_N \right) = \sum_{n=1}^{\infty} P(A_n)
P(∪n=1mAN)=n=1∑∞P(An)
Theorem 2.2
If P P P is a probability measure on ( Ω , A ) (\Omega,\mathcal{A}) (Ω,A), then:
i. We have P ( ∅ ) = 0 P( \emptyset)=0 P(∅)=0
ii. P P P is additive
↑ \uparrow ↑ 去年考的
Proof.
If in Axiom (2) we take A n = ∅ A_n= \emptyset An=∅ for all n n n, we see that the number a = P ( ∅ ) a=P( \emptyset) a=P(∅) is equal to an infinite sum of itself;
Since 0 ≤ a ≤ 1 0 \le a \le 1 0≤a≤1, this is possible only if a = 0 a=0 a=0, and we have (i).
For (ii) if suffices to apply Axiom (2) with A 1 = A A_1=A A1=A and A 2 = B A_2=B A2=B and A 3 = A 4 = ⋯ = ∅ A_3=A_4=\cdots= \emptyset A3=A4=⋯=∅, plus the fact that P ( ∅ ) = 0 P( \emptyset) =0 P(∅)=0, to obtain the additivity of P P P.
Axiom (1): P ( Ω ) = 1 P(\Omega)=1 \quad P(Ω)=1 规范性
Axiom (2):
For every countable sequence ( A n ) n ≥ 1 (A_n)_{n\ge 1} (An)n≥1 of elements of A \mathcal{A} A, pairwise disjoint (that is, A n ∩ A m = ∅ A_n \cap A_m= \emptyset An∩Am=∅ Whenever n ≠ m n \neq m n=m), one has:
P ( ∪ n = 1 ∞ A n ) = ∑ n = 1 ∞ P ( A n ) P \left( \cup_{n=1}^{ \infty} A_n \right) = \sum_{n=1}^{\infty} P \left( A_n \right) P(∪n=1∞An)=n=1∑∞P(An)
Remark
Conversely, countable additivity is not implied by additivity.
In fact, in spite of its intuitive appeal, additivity is not enough to handle the mathematical problems of the theory, even in such a simple example as tossing a coin, as we will see later.
Theorem 2.3
Let A \mathcal{A} A be a σ − a l g e b r a \sigma-algebra σ−algebra.
Suppose that P : A → [ 0 , 1 ] P: \mathcal{A} \to [0,1] P:A→[0,1] satisfies (1) and is additive.
Then the following are equivalent:
i. Axiom (2) of Definition 2.3 (i.e. P P P is countable additivity)
ii. If A n ∈ A A_n \in \mathcal{A} An∈A and A n ↓ ∅ A_n \downarrow \emptyset An↓∅, then P ( A n ) ↓ 0 P(A_n) \downarrow 0 P(An)↓0
iii. If A n ∈ A A_n \in \mathcal{A} An∈A and A n ↓ A A_n \downarrow A An↓A, then P ( A n ) ↓ P ( A ) P(A_n) \downarrow P(A) P(An)↓P(A)
iv. If A n ∈ A A_n \in \mathcal{A} An∈A and A n ↑ Ω A_n \uparrow \Omega An↑Ω, then P ( A n ) ↑ 1 P(A_n) \uparrow 1 P(An)↑1
v. If A n ∈ A A_n \in \mathcal{A} An∈A and A n ↑ A A_n \uparrow A An↑A, then P ( A n ) ↑ P ( A ) P(A_n) \uparrow P(A) P(An)↑P(A)
Proof.
The notation A n ↓ A A_n \downarrow A An↓A means that A n + 1 ⊂ A n A_{n+1} \subset A_n An+1⊂An, each n n n, and ∩ n = 1 ∞ A n = A \cap_{n=1}^{ \infty} A_n =A ∩n=1∞An=A.
The notation A n ↑ A A_n \uparrow A An↑A means that A n ⊂ A n + 1 A_n \subset A_{n+1} An⊂An+1, each n n n, and ∪ n = 1 ∞ A n = A \cup_{n=1}^{ \infty}A_n=A ∪n=1∞An=A
Note that is A n ↓ A A_n \downarrow A An↓A, then A n c ↑ A c A_n^c \uparrow A^c Anc↑Ac, and by the finite additivity axiom P ( A n c ) = 1 − P ( A n ) P(A_n^c)=1-P(A_n) P(Anc)=1−P(An).
Therefore (ii) is equivalent to (iv) and similarly (iii) is equivalent to (v).
Moreover by choosing A A A to be Ω \Omega Ω we have that (v) implies (iv)
只要证明了 (ii) ,就可以推出 (iv);只要证明了 (iii),就可以推出 (v);在 (v) 里面,取 A → Ω A\to \Omega A→Ω,则能证明 (iv)。
Suppose now that we have (iv).
Let A n ∈ A A_n \in \mathcal{A} An∈A with A n ↑ A A_n \uparrow A An↑A.
Set B n = A n ∪ A c B_n = A_n \cup A^c Bn=An∪Ac.
Then B n B_n Bn increase to Ω \Omega Ω, hence P ( B n ) P(B_n) P(Bn) increase to 1 1 1
Since A n ⊂ A A_n \subset A An⊂A we have A n ∩ A c = ∅ A_n \cap A^c= \emptyset An∩Ac=∅. Thus
1 = lim n → ∞ P ( B n ) = lim n → ∞ { P ( A n ) + P ( A c ) } 1 = \lim_{n \to \infty} P(B_n) = \lim_{n \to \infty} \left\{ P(A_n) + P(A^c) \right\} 1=n→∞limP(Bn)=n→∞lim{P(An)+P(Ac)}
hence lim n → ∞ P ( A n ) = 1 − P ( A c ) = P ( A ) \lim_{n \to \infty} P(A_n) = 1-P(A^c) = P(A) limn→∞P(An)=1−P(Ac)=P(A), and we have (v) (It means (iv) → \to → (v))
It remains to show that (i) ⇔ \Leftrightarrow ⇔ (v)
Suppose we have (v).
Let A n ∈ A A_n \in \mathcal{A} An∈A be pairwise disjoint:
\qquad that is, if n ≠ m n \neq m n=m, then A n ∩ A m = ∅ A_n \cap A_m = \emptyset An∩Am=∅
Define B n = ∪ 1 ≤ k ≤ n A k B_n = \cup_{1 \le k \le n}A_k Bn=∪1≤k≤nAk and B = ∪ k = 1 n P ( A k ) B = \cup_{k=1}^{n}P(A_k) B=∪k=1nP(Ak)
Then by the definition of a Probability Measure we have P ( B n ) = ∑ k = 1 n P ( A k ) P(B_n)= \sum_{k=1}^{n} P(A_k) P(Bn)=∑k=1nP(Ak) which increase with n n n to ∑ k = 1 n P ( A n ) \sum_{k=1}^{n} P(A_n) ∑k=1nP(An), and also P ( B n ) P(B_n) P(Bn) increases to P ( B ) P(B) P(B) by (v)
We deduce lim n → ∞ P ( B n ) = P ( B ) \lim_{n \to \infty} P(B_n) = P(B) limn→∞P(Bn)=P(B) and we have
P ( B ) = P ( ∪ n = 1 ∞ A n ) = ∑ n = 1 ∞ P ( A n ) P(B) = P( \cup_{n=1}^{ \infty} A_n) = \sum_{n=1}^{\infty} P(A_n) P(B)=P(∪n=1∞An)=n=1∑∞P(An)
and thus we have (i)
Finally assume we have (i), and we wish to establish (v)
Let A n ∈ A A_n \in \mathcal{A} An∈A, with A n A_n An increasing to A A A.
We construct a new sequence as follows:
B 1 = A 1 B 2 = A 2 − A 1 = A 2 ∩ A 1 c ⋮ B n = A n − A n − 1 \begin{aligned} B_1 &= A_1 \\ B_2 &= A_2 - A_1 = A_2 \cap A_1^c \\ \vdots & \\ B_n &= A_n - A_{n-1} \\ \end{aligned} B1B2⋮Bn=A1=A2−A1=A2∩A1c=An−An−1
Then ∪ n = 1 ∞ B n = A \cup_{n=1}^{ \infty}B_n = A ∪n=1∞Bn=A and the events ( B n ) n ≥ 1 (B_n)_{n\ge 1} (Bn)n≥1 are pairwise disjoint.
Therefore by (i) we have
P ( A ) = lim n → ∞ ∑ k = 1 n P ( B k ) P(A) = \lim_{n \to \infty} \sum_{k=1}^{n} P(B_k) P(A)=n→∞limk=1∑nP(Bk)
But also ∑ k = 1 n P ( B k ) = P ( A n ) \sum_{k=1}^{n} P(B_k)=P(A_n) ∑k=1nP(Bk)=P(An), hence we deduce lim n → ∞ P ( A n ) = P ( A ) \lim_{n \to \infty} P(A_n) = P(A) limn→∞P(An)=P(A) and we have (v)
Remark :示性函数
If A ∈ 2 Ω A \in 2^{\Omega} A∈2Ω, we define the indicator function by
1 A ( ω ) = { 1 i f ω ∈ A 0 i f ω ∉ A 1_{A}(\omega) = \left\{ \begin{array}{lll} 1 & if \ \omega \in A \\ 0 & if \ \omega \not\in A \end{array} \right. 1A(ω)={10if ω∈Aif ω∈A
We often do not explicitly write the ω \omega ω, and just write 1 A 1_{A} 1A
We can say that A n ∈ A A_{n} \in \mathcal{A} An∈A converges to A A A (we write A n → A A_n \to A An→A) if lim n → ∞ 1 A n ( ω ) = 1 A ( ω ) \lim_{n \to \infty} 1_{A_n}(\omega)=1_{A}(\omega) limn→∞1An(ω)=1A(ω) for all ω ∈ Ω \omega \in \Omega ω∈Ω
Note it also tends to A A A in the above sense.
补几个概念:
数列极限:数列 { a n } \{a_n\} {an} 收敛于 a a a 的定义:
对于每一个给定的 ε > 0 \varepsilon>0 ε>0 ,存在 N N N ,使得对满足条件 n > N n>N n>N 的每个自然数 n n n ,成立不等式 ∣ a n − a ∣ < ε \mid a_n - a \mid < \varepsilon ∣an−a∣<ε
∀ ε > 0 , ∃ N , ∀ n > N ⇒ ∣ a n − a ∣ < ε \forall \varepsilon > 0, \exists N, \forall n>N\quad \Rightarrow \quad \mid a_n-a \mid < \varepsilon ∀ε>0,∃N,∀n>N⇒∣an−a∣<ε
函数收敛:函数 F ( x ) F(x) F(x) 在 x x x 趋于 a a a 时以 A A A 为极限:
设 a , A ∈ R a,A \in \mathbb{R} a,A∈R ,函数 f f f 在点 a a a 的一个邻域中有定义,若对每一个给定的 ε > 0 \varepsilon>0 ε>0 ,存在 δ > 0 \delta>0 δ>0 ,使得当 x ∈ O δ ( a ) − { a } x \in O_{\delta}(a)-\{a\} x∈Oδ(a)−{a} (即 0 < ∣ x − a ∣ < δ 0< \mid x-a \mid <\delta 0<∣x−a∣<δ) 时,成立 ∣ f ( x ) − A ∣ < ε \mid f(x)-A \mid <\varepsilon ∣f(x)−A∣<ε
∀ ε > 0 , ∃ δ > 0 , ∀ x ∈ O δ ( a ) − { a } ⇒ ∣ f ( x ) − A ∣ < ε \forall \varepsilon > 0 , \exists \delta>0, \forall x \in O_{\delta}(a) - \{a\} \quad \Rightarrow \quad \mid f(x) - A \mid < \varepsilon ∀ε>0,∃δ>0,∀x∈Oδ(a)−{a}⇒∣f(x)−A∣<ε
函数的一致连续性:函数 f f f 在区间 I I I 上为一致连续:
如果对每一个 ε > 0 \varepsilon>0 ε>0 ,存在 δ > 0 \delta>0 δ>0 ,使得当 x ′ , x ′ ′ ∈ I x',x'' \in I x′,x′′∈I 且 ∣ x ′ − x ′ ′ ∣ < δ \mid x'-x'' \mid <\delta ∣x′−x′′∣<δ 时,成立 ∣ f ( x ′ ) − f ( x ′ ′ ) ∣ < δ \mid f(x')-f(x'') \mid < \delta ∣f(x′)−f(x′′)∣<δ
函数列的一致收敛:设一个函数列 { f n } \{f_n\} {fn} 在数集 E E E 上收敛于其极限函数 f ( x ) f(x) f(x). 称 { f n } \{f_n\} {fn} 于 E E E 上一致收敛:
如果对于每个正数 ε > 0 \varepsilon >0 ε>0,存在 N N N, 使得对每个正整数 n > N n>N n>N ,和每个 x ∈ E x \in E x∈E,均成立 ∣ f n ( x ) − f ( x ) ∣ < ε \mid f_n(x) - f(x) \mid < \varepsilon ∣fn(x)−f(x)∣<ε
f o r a n y f i x x ∈ E ∀ ε > 0 , ∃ N , ∀ n > N ⇓ ∣ f n ( x ) − f ( x ) ∣ < ε \begin{aligned} & \ for \ any \ fix \ x \in E \\ & \forall \varepsilon>0 , \exists N , \forall n> N \\ & \Downarrow \\ & \mid f_n(x) - f(x) \mid < \varepsilon \end{aligned} for any fix x∈E∀ε>0,∃N,∀n>N⇓∣fn(x)−f(x)∣<ε
Theorem 2.4
Let P be a probability measure
Let A n A_n An be a sequence of events in A \mathcal{A} A which converges to A A A
Then.
A ∈ A a n d lim n → ∞ P ( A n ) = P ( A ) A \in \mathcal{A} \quad and \quad \lim_{n \to \infty} P(A_n) = P(A) A∈Aandn→∞limP(An)=P(A)
Proof.
Let us define
lim sup n → ∞ A n = ∩ n = 1 ∞ ∪ m ≥ n A m \limsup_{n \to \infty} A_n = \cap_{n=1}^{ \infty} \cup_{m \ge n} A_m n→∞limsupAn=∩n=1∞∪m≥nAm
lim inf n → ∞ A n = ∪ n = 1 ∞ ∩ M ≥ n A m \liminf_{n \to \infty} A_n = \cup_{n=1}^{ \infty} \cap_{M \ge n} A_m n→∞liminfAn=∪n=1∞∩M≥nAm
令 B n = ∪ m ≥ n A m C n = ∩ m ≥ n A m B_n = \cup_{m\ge n} A_m \quad C_n = \cap_{m\ge n}A_m Bn=∪m≥nAmCn=∩m≥nAm
∃ ω ∈ ∪ n = 1 ∞ ∩ m ≥ n A m = ∪ n = 1 ∞ C n \exist \omega \in \cup_{n=1}^{ \infty} \cap_{m\ge n}A_m = \cup_{n=1}^{ \infty}C_n ∃ω∈∪n=1∞∩m≥nAm=∪n=1∞Cn
一定存在一个 n 0 n_0 n0 ,使得 ω ∈ C n 0 = ∩ m ≥ n 0 A m \omega \in C_{n_0}= \cap_{m\ge n_0}A_m ω∈Cn0=∩m≥n0Am
则对每个 m ≥ n 0 m\ge n_0 m≥n0 ,有 ω ∈ A m \omega \in A_m ω∈Am
则
ω ∈ ∪ m ≥ 1 A m = B 1 ω ∈ ∪ m ≥ 2 A m = B 2 ⋮ ω ∈ ∪ m ≥ n A m = B n \begin{array}{lll} \omega \in \cup_{m\ge 1}A_m = B_1 \\ \omega \in \cup_{m\ge 2}A_m = B_2 \\ \vdots \\ \omega \in \cup_{m\ge n}A_m = B_n \\ \end{array} ω∈∪m≥1Am=B1ω∈∪m≥2Am=B2⋮ω∈∪m≥nAm=Bn
则 ω ∈ ∩ n = 1 ∞ B n = ∩ n = 1 ∞ ∪ m ≥ n A m \omega \in \cap_{n=1}^{ \infty}B_n = \cap_{n=1}^{ \infty} \cup_{m\ge n}A_m ω∈∩n=1∞Bn=∩n=1∞∪m≥nAm
说明
lim inf n → ∞ A n = ∪ n = 1 ∞ ∩ M ≥ n A m ⊂ lim sup n → ∞ A m = ∩ n = 1 ∞ ∪ m ≥ n A m \liminf_{n \to \infty} A_n = \cup_{n=1}^{ \infty} \cap_{M \ge n} A_m \qquad \subset \qquad \limsup_{n \to \infty} A_m = \cap_{n=1}^{ \infty} \cup_{m \ge n} A_m n→∞liminfAn=∪n=1∞∩M≥nAm⊂n→∞limsupAm=∩n=1∞∪m≥nAm
Since A \mathcal{A} A is a σ \sigma σ-algebra, we have
lim sup n → ∞ A n ∈ A lim inf n → ∞ A n ∈ A \limsup_{n \to \infty} A_n \in \mathcal{A} \qquad \liminf_{n \to \infty} A_n \in \mathcal{A} n→∞limsupAn∈An→∞liminfAn∈A
这一步是由于上面上极限和下极限的定义,可以看出 A m A_m Am 是 σ \sigma σ-代数,则其集合的运算也是 σ \sigma σ-代数
By hypothesis A n A_n An converges to A A A, which means lim n → ∞ 1 A n = 1 A \lim_{n \to \infty} 1_{A_n}=1_{A} limn→∞1An=1A for all ω \omega ω.
A n ∈ A A_n \in \mathcal{A} An∈A converges to A A A if lim n → ∞ 1 A n ( ω ) = 1 A ( ω ) \lim_{n \to \infty} 1_{A_n}(\omega)=1_{A}(\omega) limn→∞1An(ω)=1A(ω) for all ω ∈ Ω \omega \in \Omega ω∈Ω
This is equivalent to saying that
A = lim sup n → ∞ A n = lim inf n → ∞ A n A = \limsup_{n \to \infty} A_n = \liminf_{n \to \infty} A_n A=n→∞limsupAn=n→∞liminfAn
用极限逼近的思想去证明
∀ ω ∈ A 1 A ( ω ) = 1 \forall \omega \in A \qquad 1_{A}(\omega)=1 ∀ω∈A1A(ω)=1
lim n → ∞ 1 A n ( ω ) = 1 A ( ω ) = 1 \lim_{n \to \infty} 1_{A_n}(\omega)= 1_{A}(\omega)=1 n→∞lim1An(ω)=1A(ω)=1
∃ n 0 s . t . n ≥ n 0 1 A n ( ω ) = 1 s o . ω ∈ A n , n ≥ n 0 ω ∈ ∩ n ≥ n 0 A n ⊂ ∪ n 0 ≥ 1 ∩ n ≥ n 0 A n = lim inf n → ∞ A n ⇒ A ⊂ lim inf n → ∞ A n \begin{array}{llll} \exists n_0 \quad s.t. \quad n \ge n_0 \quad 1_{A_n}(\omega) =1 \\ so. \quad \omega \in A_n, n\ge n_0 \\ \omega \in \cap_{n\ge n_0} A_n \subset \cup_{n_0 \ge 1} \cap_{n\ge n_0} A_n = \liminf_{n \to \infty} A_n \\ \\ \Rightarrow A \subset \liminf_{n \to \infty} A_n \end{array} ∃n0s.t.n≥n01An(ω)=1so.ω∈An,n≥n0ω∈∩n≥n0An⊂∪n0≥1∩n≥n0An=n→∞liminfAn⇒A⊂n→∞liminfAn
∀ ω ∈ A c 1 A n ( ω ) = 0 \forall \omega \in A^{c} \qquad 1_{A_n}(\omega)=0 ∀ω∈Ac1An(ω)=0
lim n → ∞ 1 A n ( ω ) = 1 A ( ω ) = 0 \lim_{n \to \infty}1_{A_n}(\omega)=1_{A}(\omega)=0 n→∞lim1An(ω)=1A(ω)=0
∃ n 1 s . t . n ≥ n 1 1 A n ( ω ) = 0 s o . ω ∈ A n c , n ≥ n 1 ω ∈ ∩ n ≥ n 0 A n c ⊂ ∪ n 0 ≥ 1 ∩ n ≥ n 0 A n c = ( ∩ n 1 ≥ 1 ∪ n ≥ n 1 A n ) c = ( lim sup n → ∞ A n ) c s o . A c ⊂ ( lim sup n → ∞ A n ) c ⇒ lim inf n → ∞ A n ⊂ A \begin{array}{llll} \exists n_1 \quad s.t. \quad n \ge n_1 \quad 1_{A_n}(\omega) =0 \\ so. \quad \omega \in A_n^c, n\ge n_1 \\ \omega \in \cap_{n\ge n_0} A_n^c \subset \cup_{n_0 \ge 1} \cap_{n\ge n_0} A_n^c = \left( \cap_{n_1 \ge 1} \cup_{n\ge n_1}A_n \right)^c \\ \qquad \qquad \qquad \qquad \qquad \qquad \qquad = \left(\limsup_{n \to \infty} A_n \right)^c\\ \\ so.\quad A^c \subset \left(\limsup_{n \to \infty} A_n \right)^c \\ \\ \Rightarrow \liminf_{n \to \infty} A_n \subset A \end{array} ∃n1s.t.n≥n11An(ω)=0so.ω∈Anc,n≥n1ω∈∩n≥n0Anc⊂∪n0≥1∩n≥n0Anc=(∩n1≥1∪n≥n1An)c=(n→∞limsupAn)cso.Ac⊂(n→∞limsupAn)c⇒n→∞liminfAn⊂A
lim sup A n ⊂ A ⊂ lim inf A n \limsup A_n \subset A \subset \liminf A_n limsupAn⊂A⊂liminfAn
so
A = lim sup n → ∞ A n = lim inf n → ∞ A n A = \limsup_{n \to \infty} A_n = \liminf_{n \to \infty} A_n A=n→∞limsupAn=n→∞liminfAn
Therefore A ∈ A A \in \mathcal{A} A∈A
Now, let B n = ∪ m ≥ n A m C n = ∩ m ≥ n A m B_n = \cup_{m\ge n} A_m \quad C_n = \cap_{m\ge n}A_m Bn=∪m≥nAmCn=∩m≥nAm
Then
B n B_n Bn decreases to A A A : B n B_n Bn 是并集,随着 n n n 的增大,并集的个数减小,所以 B n B_n Bn 会减小到 A A A
C n C_n Cn increases to A A A : C n C_n Cn 是交集,随着 n n n 的增大,交集的个数减小,所以 C n C_n Cn 会增大到 A A A
注意这里我的写法和书上不一致,主要是为了和前面的保持一致
Thus (by theorem 2.3)
lim n → ∞ P ( B n ) = lim n → ∞ P ( C n ) = P ( A ) \lim_{n \to \infty}P(B_n)=\lim_{n \to \infty} P(C_n)=P(A) n→∞limP(Bn)=n→∞limP(Cn)=P(A)
However C n ⊂ A n ⊂ B n C_n \subset A_n \subset B_n Cn⊂An⊂Bn , therefore
P ( C n ) ≤ P ( A n ) ≤ P ( B n ) P(C_n) \le P(A_n) \le P(B_n) P(Cn)≤P(An)≤P(Bn)
So.
lim n → ∞ P ( A n ) = P ( A ) \lim_{n \to \infty} P(A_n) = P(A) n→∞limP(An)=P(A)
Exercise 2.6
Let A \mathcal{A} A be a σ \sigma σ-algebra of subsets of Ω \Omega Ω
Let B ∈ A B \in \mathcal{A} B∈A
- Show that F = { A ∩ B : A ∈ A } \mathcal{F}= \left\{ A \cap B : A \in \mathcal{A} \right\} F={A∩B:A∈A} is a σ \sigma σ-algebra of subsets of B B B.
- Is it still true when B B B is a subset of Ω \Omega Ω that does not belong to A \mathcal{A} A
Proof.
- 证明空集和全集(此时全集为 B B B )在 F \mathcal{F} F 上
∅ = ∅ ∩ B ∅ ∈ A ⇒ ∅ ∈ F B = B ∩ B B ∈ A ⇒ B ∈ F \begin{array}{rlll} \emptyset= \emptyset \cap B & \emptyset \in \mathcal{A} & \Rightarrow \emptyset \in \mathcal{F} \\ B = B \cap B & B \in \mathcal{A} & \Rightarrow B \in \mathcal{F} \end{array} ∅=∅∩BB=B∩B∅∈AB∈A⇒∅∈F⇒B∈F
- 证明对于任意的集合 C C C 在 F \mathcal{F} F 里面,它的补集也在 F \mathcal{F} F 里面
$$
\begin{array}{lll}
\forall C \in \mathcal{F} & \exists A \in \mathcal{A} \quad s.t. C = AB \
&
\begin{aligned}
B-C = B \cap C^c &= B \cap \left( A \cap B \right)^c \
&= B \cap \left( A^c + B^c \right) \
&= BA^c + BB^c \
&= BA^c \
\end{aligned}
\
& \because A \in \mathcal{A} \qquad \therefore A^c \in \mathcal{A} \\
& \therefore C^c=B-C \in \mathcal{F}
\end{array}
$$
- 证明对于可数并和可数交,其仍在 F \mathcal{F} F 里面
i f C n ∈ F ∃ A n ∈ A s . t . C n = A n B ∪ n = 1 ∞ C n = ∪ n = 1 ∞ ( A n B ) = ( ∪ n = 1 ∞ A n ) B ∴ ∪ n = 1 ∞ A n ∈ A ∴ ∪ n = 1 ∞ C n ∈ F \begin{array}{lll} if \ C_n \in \mathcal{F} & \exists A_n \in \mathcal{A} \quad s.t. C_n = A_n B \\ & \cup_{n=1}^{ \infty} C_n = \cup_{n=1}^{ \infty} (A_n B) = \left( \cup_{n=1}^{ \infty} A_n \right)B \\ \\ & \therefore \cup_{n=1}^{ \infty} A_n \in \mathcal{A} \\ & \therefore \cup_{n=1}^{ \infty} C_n \in \mathcal{F} \end{array} if Cn∈F∃An∈As.t.Cn=AnB∪n=1∞Cn=∪n=1∞(AnB)=(∪n=1∞An)B∴∪n=1∞An∈A∴∪n=1∞Cn∈F
Exercise 2.7
Exercise 以前讲过
Chapter 3 以后补。。。