Chapter 1 (Sample Space and Probability): Probabilistic models (概率模型)

本文为 $Introduction$ $to$ $Probability$ 的读书笔记

Probability Models 概率模型

• A probabilistic model is a mathematical description of an uncertain situation. Its two main ingredients are listed below and are visualized in Fig. 1.2.
  

Sample space: 样本空间

Sample Spaces and Events 样本空间和事件

• Every probabilistic model involves an underlying process, called the experiment, that will produce exactly one out of several possible outcomes.
  • It is important to note that in our formulation of a probabilistic model, there is only one experiment. So, three tosses of a coin constitute (组成) a single experiment rather than three experiments.
• The set of all possible outcomes is called the sample space of the experiment, and is denoted by $\Omega$.
• A subset of the sample space, that is, a collection of possible outcomes, is called an event (事件 / 随机事件).
  • 设 $A$ 和 $B$ 为两个随机事件，事件 $A\cup B$ 称为 $A$ 和 $B$ 的和事件，也记为 $A+B$
  • 事件 $A\cap B$ 称为 $A$ 和 $B$ 的积事件，也记为 $AB$
  • 事件 $A-B$ 称为 $A$ 和 $B$ 的差事件；注意到 $A-B=A\bar{B}$
  • 事件 $\Omega -A$ 称为 $A$ 的补事件，也记为 $\bar{A}$ 或 $A^C$

---

• 补充：De Morgan’s laws (德摩根律)
  $$(\underset{n}{\cup }{S}_n{)}^c=\underset{n}{\cap }{S}_n^c,(\underset{n}{\cap }{S}_n{)}^c=\underset{n}{\cup }{S}_n^c$$PROOF
  [Hint: if $x\in (\underset{n}{\cup }{S}_n{)}^c$, then $x\in \underset{n}{\cap }{S}_n^c$]

Sequential Models 序贯模型

• Many experiments have an inherently sequential character; for example, tossing a coin three times. It is then often useful to describe the experiment and the associated sample space by means of a tree-based sequential description(序贯树形图), as in Fig. 1.3.
  

die: 骰子

Note that every node of the tree can be identified with an event. For example, the node labeled by a 1 can be identified with the event {(1, 1), (1, 2). (1, 3), (1, 4) } that the result of the first roll is 1.

Probability Laws 概率律

• The probability law assigns to every event $A$. a number $P(A)$, called the probability of $A$. satisfying the following axioms.
  
• Inference (推论):
  $$\begin{gathered} 1=P(\Omega )=P(\Omega \cup \varnothing )=P(\Omega )+P(\varnothing )=1+P(\varnothing ) \\ \therefore P(\varnothing )=0 \end{gathered}$$

Properties of Probability Laws

• 减法公式：$P(A-B)=P(A\bar{B})=P(A)-P(AB)$
• 加法公式：$P(A\cup B)=P(A)+P(B)-P(AB)$
  • 推广：$P(A\cup B\cup C)=P(A)+P(B)+P(C)-P(AB)-P(AC)-P(BC)+P(ABC)$

---

Problem 9

A partition of the sample space $\Omega$ is a collection of disjoint events $S_1,...,S_n$ such that $\Omega ={\cup }_{i=1}^n{S}_i$.

• (a) Show that for any event $A$, we have
  $$P(A)=\sum _{i=1}^{n}P(A\cap S_i)$$
• (b) Use part (a) to show that for any events $A,B$, and $C$, we have
  $$P(A)=P(A\cap B)+P(A\cap C)+P(A\cap B^C\cap C^C)-P(A\cap B\cap C)$$

---

Problem 10.

Show the formula
$$P((A\cap B^C)\cup (A^C\cap B))=P(A)+P(B)-2P(A\cap B),$$which gives the probability that exactly one of the events $A$ and $B$ will occur.

SOLUTION
$$\begin{array}{rl}& P((A\cap B^C)\cup (A^C\cap B))\\ =& P(A\cap B^C)+P(A^C\cap B)\\ =& P(A)-P(A\cap B)+P(B)-P(A\cap B)\end{array}$$

---

Problem 11 Bonferroni’s inequality (邦费罗尼不等式)

• (a) Prove that for any two events $A$ and $B$, we have
  $$P(A\cap B)\ge P(A)+P(B)-1$$

SOLUTION

• (a) We have $P(A\cup B)=P(A)+P(B)-P(A\cap B)$ and $P(A\cup B)\le 1$. which implies part (a).

也可以这么做： $1+P(A\cap B)=2P(A\cap B)+P(A\cap B^C)+P(A^C\cap B)+P(A^C\cap B^C)=(P(A\cap B)+P(A\cap B^C))+(P(A\cap B)+P(A^C\cap B))+P(A^C\cap B^C)=P(A)+P(B)+P(A^C\cap B^C)\ge P(A)+P(B)$

---

Problem 13. Continuity property of probabilities (概率的连续性)

• $(a)$ Let $A_1,A_2,...$. be an infinite sequence of events, which is “monotonically increasing,” meaning that $A_n\subset A_{n+1}$ for every $n$. Let $A={\cup }_{n=1}^{\infty }{A}_n$. Show that $P(A)=li{m}_{n\to \infty }P(A_n)$
  [Hint: Express the event $A$ as a union of countably many disjoint sets.]
• $(b)$ Suppose now that the events are “monotonically decreasing,” i.e., $A_{n+1}\subset A_n$ for every $n$. Let $A={\cap }_{n=1}^{\infty }{A}_n$ . Show that $P(A)=li{m}_{n\to +\infty }P(A_n)$.
• $(c)$ Consider a probabilistic model whose sample space is the real line. Show that
  $$\begin{gathered} P([0,\infty ))=li{m}_{n\to \infty }P([0,n]) \\ li{m}_{n\to \infty }P([n,\infty ))=0 \end{gathered}$$

SOLUTION

• (a) Let $B_1=A_1$ and, for $n\ge 2,B_n=A_n\cap A_{n-1}^C$ . The events $B_n$ are disjoint, and we have ${\cup }_{k=1}^n{B}_k=A_n$, and ${\cup }_{k=1}^{\infty }{B}_k=A$.
  $$P(A)=\sum _{k=1}^{\infty }P(B_k)=P({\cup }_{k=1}^{\infty }{B}_k)=li{m}_{n\to \infty }P({\cup }_{k=1}^n{B}_k)=li{m}_{n\to \infty }P(A_n)$$
• (b) [Hint: Apply the result of part (a) to the complements of the events $P(A^C)$.]
• ($c$) For the first equality, use the result frorn part (a) with $A_n=[0,n]$ and $A=[0,\infty )$. For the second, use the result from part (b) with $A_n=[n,\infty )$ and $A=\varnothing$.

Discrete Models 离散模型

离散概率律

Note that we are using here the simpler notation $P(s_i)$ to denote the probability of the event { s i } \{s_i\} {si​} , instead of the more precise P ( { s i } ) P(\{ s_i\}) P({si​}).

---

离散均匀概率律 (古典概型)

• In the special case where the probabilities $P(s_1)$, … , $P(s_n)$ are all the same (by necessity equal to $1/n$), we obtain the following.
  

Continuous Models 连续模型

• Probabilistic models with continuous sample spaces differ from their discrete counterparts in that the probabilities of the single-element events may not be sufficient to characterize the probability law.
• 几何概型：连续均匀概率律

Example 1.4

• A wheel of fortune (幸运轮) is continuously calibrated from 0 to 1, so the possible outcomes of an experiment consisting of a single spin are the numbers in the interval $n=[0,1]$. Assuming a fair wheel, it is appropriate to consider all outcomes equally likely, but what is the probability of the event consisting of a single element? It cannot be positive, because then, using the additivity axiom, it would follow that events with a sufficiently large number of elements would have probability larger than 1. Therefore, the probability of any event that consists of a single element must be 0.
• In this example, it makes sense to assign probability $b-a$ to any subinterval $[a,b]$ of $[0,1]$, and to calculate the probability of a more complicated set by evaluating its “length”.
• The legitimacy of using length as a probability law hinges on the fact that the unit interval has an uncountably infinite number of elements.