📚 Bayes' Theorem | 贝叶斯定理
Bayes' theorem is a cornerstone of probability theory, providing a rigorous method for updating the likelihood of a hypothesis when new evidence becomes available. Named after Thomas Bayes, it forms the basis of modern statistical inference and is deeply embedded in the IB Mathematics Analysis and Approaches syllabus. Mastering this theorem empowers you to reason about uncertainty in a systematic way, whether you are interpreting medical test results or analysing data-driven decisions.
贝叶斯定理是概率论的基石，它提供了一种在新证据出现时更新假设可能性的严格方法。该定理以托马斯·贝叶斯命名，构成了现代统计推断的基础，并深深嵌入IB数学分析与方法的课程大纲中。掌握这一定理能够让你系统性地对不确定性进行推理，无论是解读医学检测结果还是分析数据驱动的决策。
1. Introduction to Conditional Probability | 条件概率介绍
Conditional probability measures the chance of an event A occurring given that another event B has already occurred. It is denoted as P(A | B) and defined as the ratio of the probability of both events happening to the probability of the conditioning event, provided P(B) > 0.
条件概率衡量在另一个事件B已经发生的条件下，事件A发生的概率。它记为P(A | B)，定义为两事件同时发生的概率与条件事件概率之比，前提是P(B) > 0。
P(A | B) = P(A ∩ B) / P(B)
This definition is fundamental. It essentially restricts the sample space to event B and then finds how much of A lies within that restricted space. Without a solid grasp of conditional probability, Bayes' theorem cannot be understood.
这一定义是根本性的。它本质上将样本空间限制到事件B，然后确定A有多少位于该限制空间内。如果对条件概率没有扎实的掌握，就无法理解贝叶斯定理。
2. The Law of Total Probability | 全概率公式
Before we can apply Bayes' theorem, we often need the law of total probability. This law expresses the probability of an event B as a weighted sum of conditional probabilities, based on a partition of the sample space. If events A₁, A₂, ..., Aₙ form a partition (mutually exclusive and exhaustive), then:
在应用贝叶斯定理之前，我们经常需要全概率公式。该定律将事件B的概率表示为基于样本空间划分的条件概率的加权和。如果事件A₁, A₂, ..., Aₙ构成一个划分（互斥且完备），则：
P(B) = Σ P(B | Aᵢ) × P(Aᵢ)
In the simplest case with an event A and its complement A', the law becomes P(B) = P(B | A)P(A) + P(B | A')P(A'). This formula is used as the denominator in Bayes' hat the posterior probability is properly normalised by all the ways evidence B can occur.
在最简单的情况下，涉及事件A及其补集A'，该公式变为P(B) = P(B | A)P(A) + P(B | A')P(A')。这个公式作为贝叶斯定理的分母，确保后验概率通过证据B可能出现的所有方式进行了正确的归一化。
3. Deriving Bayes' Theorem | 贝叶斯定理的推导
Bayes' theorem emerges elegantly from the definition of conditional probability. We know that P(A | B) = P(A ∩ ). Also, the joint probability can be written using the other conditional probability: P(B | A) = P(A ∩ B) / P(A), which gives P(A ∩ B) = P(B | A)P(A). Substituting this into the numerator yields:
贝叶斯定理优雅地从条件概率的定义中产生。我们知道P(A | B) = P(A ∩ B) / P(B)。同时，联合概率也可以用另一个条件概率表示：P(B | A) = P(A ∩ B) / P(A)，得到P(A ∩ B) = P(B | A)P(A)。将其代入分子得到：
P(A | B) = P(B | A) × P(A) / P(B)
If we then expand the denominator using the law of total probability for partition {A, A'}, we obtain the most commonly used form in IB exam problems:
如果我们再用划分{A, A'}的全概率公式展开分母，就得到IB考试中最常用的形式：
P(A | B) = P(B | A) × P(A) / [P(B | A)P(A) + P(B | A')P(A')]
This derivation highlights an essential point: Bayes' theorem is not an isolated formula but a direct consequence of the basic rules of probability. It simply reverses the direction of conditioning, allowing us to go from P(B | A) to P(A | B).
这一推导突出了一个要点：贝叶斯定理并非孤立的公式，而是概率基本规则的直接结果。它只是反转了条件作用的方向，使得我们能从P(B | A)得到P(A | B)。
4. Understanding the Components | 理解定理的各个部分
Each term in Bayes' theorem has a specific name and interpretation. The prior probability P(A) represents our initial belief about event A before observing any evidence. The likelihood P(B | A) is the probability of observing evidence B given that A is true. The marginal probability P(B), often called the evidence, acts as a normalising constant. Finally, the posterior probability P(A | B) is our updated belief after taking the evidence into account.
贝叶斯定理中的每一项都有特定的名称和解释。先验概率P(A)代表在观察到任何证据之前，我们对事件A的初始信念。似然概率P(B | A)是在A为真的条件下观察到证据B的概率。边际概率P(B)，常被称为证据，充当归一化常数。最后，后验概率P(A | B)是我们在考虑证据后更新的信念。
ce. This structure is the heart of Bayesian inference, and every IB question testing Bayes' theorem requires you to correctly identify these three components.
一个有用的记忆方法是：后验 = (似然 × 先验) / 证据。这个结构是贝叶斯推断的核心，每一道考查贝叶斯定理的IB题目都要求你正确识别这三个组成部分。
5. Worked Example 1: Diagnostic Testing | 例题1：诊断测试
Suppose a disease affects 1% of a population. A test for the disease has a sensitivity (true positive rate) of 95% and a specificity (true negative rate) of 90%. If a randomly selected person tests positive, what is the probability they actually have the disease? Let D be the event of having the disease and '+' be a positive test result.
假设某疾病影响1%的人口。针对该疾病的检测具有95%的灵敏度（真阳性率）和90%的特异度（真阴性率）。如果一名随机选择的人检测呈阳性，那么他实际患病的概率是多少？设D为患病事件，'+'为检测阳性结果。
We are given: P(D) = 0.01 (prior), P(+ | D) = 0.95 (sensitivity), P(+ | D') = 1 - 0.90 = 0.10 (false positive rate). The required probability is P(D | +). Applying Bayes' theorem:
我们已知：P(D) = 0.01（先验），P(+ | D) = 0.95（灵敏度），P(+ | D') = 1 - 0.90 = 0.10（假阳性率）。所求概率为P(D | +)。应用贝叶斯定理：
P(D | +) = [P(+ | D)P(D)] / [P(+ | D)P(D) + P(+ | D')P(D')]
Substituting values: P(D | .01) / (0.95 × 0.01 + 0.10 × 0.99) = 0.0095 / (0.0095 + 0.099) = 0.0095 / 0.1085 ≈ 0.0876.
代入数值：P(D | 0.01) / (0.95 × 0.01 + 0.10 × 0.99) = 0.0095 / (0.0095 + 0.099) = 0.0095 / 0.1085 ≈ 0.0876。
The posterior probability is only about 8.76%. This counterintuitive result underscores how a low prior probability can strongly influence the posterior, even when the test appears highly accurate. Many IB problems are designed to test this insight.
后验概率仅约为8.76%。这个反直觉的结果强调了即使检测看似非常精准，较低的先验概率也能强烈影响后验概率。很多IB题目正是为了考查这一洞见而设计的。
6. Worked Example 2: Spam Filter | 例题2：垃圾邮件过滤
An email filtering system classifies messages as spam or regular. Historically, 40% of all emails are spam. The word "free" appears in 80% of spam messages but only in 10% of regular emails. If an incoming email contains the word "free", what is the probability it is spam? Let S denote spam and F denote the presence of "free".
某邮件过滤系统将信息分类为垃圾邮件或普通邮件。历史上，所有邮件中有40%是垃圾邮件。单词"free"出现在80%的垃圾邮件中，但在普通邮件中只出现10%。如果一封收到的邮件包含单词"free"，它是垃圾邮件的概率是多少？设S表示垃圾邮件，F表示包含"free"。
Given: P(S) = 0.4, P(F | S) = 0.8, P(F | S') = 0.1. Using Bayes' theorem:
已知：P(S) = 0.4，P(F | S) = 0.8，P(F | S') = 0.1。使用贝叶斯定理：
P(S | F) = [P(F | S)P(S)] / [P(F | S)P(S) + P(F | S')P(S')]
P(S') is 0.6. So P(S | 0.8 × 0.4 + 0.1 × 0.6) = 0.32 / (0.32 + 0.06) = 0.32 / 0.38 ≈ 0.8421. Thus, given the word "free", there is a roughly 84.2% chance the email is spam.
P(S')为0.6。因此P(S | F) = (0.8 × 0.4) / (0.8 × 0.4 + 0.1 × 0.6) = 0.32 / (0.32 + 0.06) = 0.32 / 0.38 ≈ 0.8421。因此，给定单词"free"，该邮件有大约84.2%的概率是垃圾邮件。
7. Using Tree Diagrams | 使用树状图
Tree diagrams are an exceptionally useful tool for visualising Bayes' theorem problems. Start by branching for the prior probabilities of the hypothesis (e.g., disease or spam). Then from each branch, add the conditional probabilities for the evidence. The joint probabilities for each path are found by multiplying along the branches. The posterior probability is then the joint probability of the path of interest divided by the sum of all paths that lead to the observed evidence.
树状图是可视化贝叶斯定理问题的一个极其有用的工具。首先为假设的先验概率（如患病或垃圾邮件）画出分支。然后从每个分支出发，添加关于证据的条件概率。每条路径的联合概率通过沿分支相乘得到。后验概率就是关注路径的联合概率，除以所有导致已观察证据的路径概率之和。
In the medical test example, the tree would show two initial branches: D (0.01) and D' (0.99). From D, branch to '+' (0.95) and '–' (0.05). From D', branch to '+' (0.10) and '–' (0.90). P(D ∩ +) = 0.01 × 0.95 = 0.0095, while P(D' ∩ od reduces misidentification of the denominator and is highly recommended by IB examiners.
在医学检测例子中，树状图会显示两条初始分支：D (0.01) 和 D' (0.99)。从D出发，分支到'+' (0.95) 和 '–' (0.05)。从D'出发，分支到'+' (0.10) 和 '–' (0.90)。P(D ∩ +) = 0.01 × 0.95 = 0.0095，而P(D' ∩ +) = 0.99 × 0.10 = 0.0990。后验概率为0.0095/(0.0095+0.0990)。此方法可减少分母的误判，IB考官强烈推荐使用。
8. Common Pitfalls and Misconceptions | 常见陷阱与误解
One of the most frequent errors is confusing P(A | B) with P(B | ic scenario. Another common mistake is using the wrong denominator by omitting some terms from the law of total probability, particularly when there are more than two hypotheses. Ignoring the prior probability entirely and relying solely on the likelihood is known as the base rate fallacy, a classic trap in Bayesian reasoning.
最常见的错误之一是混淆P(A | B)和P(B | A)。学生常认为它们相等或可互换，但除非在对称情形下P(A) = P(B)，否则并非如此。另一个常见错误是分母出错，漏掉了全概率公式中的某些项，尤其是当存在两个以上假设时。完全忽略先验概率而仅依赖似然概率被称为基础比率谬误，是贝叶斯推理中的经典陷阱。
Additionally, students sometimes misinterpret the complement correctly. If specificity is given as 90%, the false positive rate is 10%, not 90%. Carefully reading the wording and defining events explicitly at the start of the solution shields against such slips. Always double-check that probabilities sum to 1 along each branch of a tree diagram.
此外，学生有时会错误地解读补集。如果特异度给出为90%，假阳性率是10%，而不是90%。在解题开始时仔细阅读措辞并明确定义事件，可以避免此类失误。务必反复检查树状图每条分支上的概率是否总和为1。
9. IB Exam Tips | IB考试技巧
In IB Mathematics: Analysis and Approaches, Bayes' theorem questions often appear in Paper 1 (non-calculator) or Paper 2 (calculator). For Paper 1, you must be able to set up the formula symbolically and simplify exact fractions. For Paper 2, precise decimal answers are expected, but you should still show the unsimplified expression to earn method marks. Always begin by defining your events with clear notation, such as 'Let D be the event...' and write down the given probabilities in terms of these events.
在IB数学：分析与方法的考试中，贝叶斯定理的题目常出现在试卷一（无计算器）或试卷二（可使用计算器）。对于试卷一，你必须能够使用符号建立公式并化简为精确分数。对于试卷二，虽然需要精确的小数答案，但仍应展示未化简的表达式以获取方法分。始终以清晰的符号定义事件开始，如 "Let D be the event..."，并用这些事件写出给定的概率。
Draw a tree diagram even if the problem seems simple; it serves as a visual check. Then write the Bayes' theorem formula with the relevant terms substituted before performing any calculations. If the answer seems counterintuitive, relate it back to the prior probability to confirm it makes sense. These exam habits, consistently applied, will significantly boost your marks.
即使问题看似简单，也要画出树状图；它可作为视觉上的核对。然后在进行任何计算之前，写出代入相关项后的贝叶斯定理公式。如果答案看似反直觉，将其与先验概率联系起来以确认其合理性。这些一贯应用的考试习惯将显著提高你的分数。
10. Real-World Applications | 现实世界应用
Bayes' theorem extends far beyond textbook examples. In medicine, it underpins the interpretation of diagnostic tests and screening programmes. In machine learning, naive Bayes classifiers use this theorem to categorise texts or images by updating the probability of a class given the features observed. In legal contexts, Bayesian reasoning helps evaluate the weight of forensic evidence, though its use in courts remains controversial.
贝叶斯定理的运用远远超出教科书中的例子。在医学领域，它支撑着诊断检测和筛查项目的解读。在机器学习中，朴素贝叶斯分类器利用该定理，通过给定观察到的特征更新类别的概率，从而对文本或图像进行分类。在法律背景下，贝叶斯推理有助于评估法医证据的权重，尽管它在法庭上的使用仍存争议。
Even in everyday decision-making, we subconsciously apply Bayesian-like updating when we revise our opinions after encountering new information. The COVID-19 pandemic made the concepts of sensitivity, specificity and false positive rates highly relevant to the public. Understanding Bayes' theorem equips you not just for an exam, but for a lifetime of critical thinking in a data-rich world.
即使在日常决策中，当我们遇到新信息后修正观点时，我们也会下意识地运用类似贝叶斯的更新过程。新冠疫情使得灵敏度、特异度和假阳性率等概念与公众高度相关。理解贝叶斯定理不仅是为了考试，更是为了在数据丰富的世界中拥有终身批判性思维的能力。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图