Notes on Probability Essentials - 2 - Conditional Probability and Independence
Probability
百年歌自苦,未见有知音。
——杜甫,《南征》
Definition 1 (a) Two events A and B are independent if P(A∩B)=P(A)P(B).
(b) A (possibly infinite) collection of events (Ai)i∈I is an independent collection if for every finite subset J of I one has
P(∩i∈JAi)=∏i∈JP(Ai)
The collection
(Ai)i∈J is often said to be mutually independent.
Theorem 1 If A and B are independent, so also are A and Bc, Ac and B, Ac and Bc
Proof: For A and Bc,
P(A∩Bc)=P(A)−P(A∩B)=P(A)−P(A)P(B)=P(A)(1−P(B))=P(A)P(Bc)
For
Ac and
B,
P(Ac∩B)=P(B)−P(A∩B)=P(B)−P(A)P(B)=P(B)(1−P(A))=P(Ac)P(B)
For
Ac and
Bc,
P(Ac∩Bc)=P(Ac)−P(Ac∩B)=P(Ac)−P(Ac)P(B)=P(Ac)(1−P(B))=(1−P(A))(1−P(B))=P(Ac)P(Bc)
Definition 2 Let A,B be events, P(B)>0, the conditional probability of A given B is P(A|B)=P(A∩B)/P(B).
Remark. 在这里谈一下对条件概率的感性认识。 许多时候,一个事情的发生多多少少会影响另外一件事情发生的可能性。
那么计算方法为什么是P(A|B)=P(A∩B)P(B)呢?P(A∩B)代表两个事件同时发生,现在确实是同时发生了,但是原先B的发生并不是必然的,例如原先只有5%的可能性会发生B,但是现在这个5%已经确定必然发生了,变成了100%,放大了20倍,那么这样一来A发生的概率也就跟着“等比例放大了20倍”。
若这样凭空感受太抽象,那不妨举个著名的例子——(Monty Hall Problem),这可能是历史上最有争议的概率问题,问题看似简单但正确答案如此有悖常理以至于很多人不能接受。问题描述如下——
Theorem 2Proof: Part(1) seems to be a direct result from Definition 1 and Definition 2.
Part(2), define Q(A)=P(A|B), with B fixed. We must show Q satisfies the definition of a probability measure.
Q(Ω)=P(Ω|B)=P(Ω∩B)P(B)=P(B)P(B)=1
If
(An)n≥1 is a sequence of elements of
A which are pairwise disjoint, then
Q(∪∞n=1An)=P(∪∞n=1An|B)=P(∪∞n=1(An∩B))P(B)
also the sequence
(An∩B)n≥1 is pairwise disjoint as well; thus
=∑n=1∞P(An∩B)P(B)=∑n=1∞P(An|B)=∑n=1∞Q(An)
Theorem 3 If A1,...,An∈A and if P(A1∩...∩An−1)>0, then
P(A1∩...∩An)=P(A1)P(A2|A1)P(A3|A1∩A2)...P(An|A1∩...∩An−1)
Proof.(draft) By Induction. For n=2, the theorem is simply Definition 2. Suppose the theorem holds for n−1 events. Let B=A1∩...∩An−1...
Theorem 4 (Partition Equation). Let (En)n≥1 be a finite or countable partition of Ω. Then if A∈A,
P(A)=∑nP(A|En)P(En)
Theorem 5 (Bayes' Theorem) Let (En) be a finite or countable partition of Ω and suppose P(A)>0. Then
P(En|A)=P(A|En)P(En)∑mP(A|Em)P(Em)
Note: 贝叶斯定理的表述极其简单,等式右侧分子为
P(A∩En),分母为
P(A),基本就是条件概率公式遇到互斥事件(
Em)时的一种应用而已。可是不难看出这样的一种“展开形式”给
P(En|A)和
P(A|En)这两个量之间建立了一种关联,于是贝叶斯定理成了一个具有里程碑意义的重要定理。
点击这里查看贝叶斯定理的一个应用:
《用贝叶斯定理来讨论“医疗诊断的可靠性到底有多少”》