📚 Poisson Distributions | 泊松分布
The Poisson distribution is a discrete probability distribution that expresses the probability of a given number of events occurring in a fixed interval of time or space, provided the events happen with a known constant mean rate and independently of the time since the last event. In Edexcel A-Level Mathematics, it is an essential topic within the Statistics component, applied to model rare or random occurrences such as phone calls at a call centre per minute, radioactive decays per second, or flaws in a length of wire.
泊松分布是一种离散概率分布，用来描述在固定的时间或空间区间内某事件发生的次数，条件是该事件以已知的恒定平均速率发生，且各事件独立发生。在Edexcel A-Level数学中，泊松分布是统计学部分的重要课题，常用于对稀有或随机事件建模，如每分钟呼叫中心接到的电话数、每秒放射性衰变次数，或导线单位长度上的瑕疵数。
1. Introduction to the Poisson Distribution | 泊松分布简介
The Poisson distribution arises when we count the number of occurrences of an event over a given continuous domain – typically time, area, volume, or length – under a stable average rate. It is named after the French mathematician Siméon Denis Poisson. The key parameter, denoted by the Greek letter λ (lambda), represents the expected number of occurrences in that interval. A random variable X that follows a Poisson distribution is written as X ~ Po(λ).
泊松分布适用于对在给定连续域（通常是时间、面积、体积或长度）内事件发生次数的计数，前提是平均发生率保持稳定。它以法国数学家西莫恩·德尼·泊松命名。关键参数用希腊字母 λ 表示，代表该区间内事件发生的期望次数。服从泊松分布的随机变量 X 记为 X ~ Po(λ)。
The distribution is discrete because the outcome is a non‑negative integer count: 0, 1, 2, ... . Typical textbook examples include the number of misprints per page of a book, the number of customers arriving at a bank in a 10‑minute window, or the number of earthquakes in a region over a year. Its simplicity and usefulness lie in the fact that a single parameter λ fully defines the shape and scale of the distribution.
该分布是离散的，因为结果是 0,1,2,… 等非负整数计数。典型的教材示例包括一本书每页的印刷错误数、银行10分钟内到达的顾客人数，或某地区一年的地震次数。泊松分布简洁实用的原因在于：仅需一个参数 λ 便可完整确定分布的形状与规模。
2. Conditions for a Poisson Distribution | 泊松分布的适用条件
For a phenomenon to be modelled well by a Poisson distribution, four main conditions must be satisfied. First, events must occur singly – the probability of two or more events happening at exactly the same instant in a sufficiently small interval is negligible. Second, events must be independent: the occurrence of one event does not affect the probability of another event occurring.
要使用泊松分布对现象进行合理建模，必须满足四个主要条件。第一，事件必须单独发生——在极短的区间内，同时发生两个或以上事件的概率可忽略不计。第二，事件间必须独立：一个事件的发生不影响另一个事件发生的概率。
Third, the average rate at which events occur, λ, must be constant throughout the interval. This means the process is stationary. Fourth, events occur at random – there is no pattern or clustering that makes some sub‑intervals more likely than others. In practice, these conditions are often assumed rather than proved, but verifying them is part of a good statistical analysis.
第三，事件发生的平均速率 λ 在整个区间内必须恒定。这意味着过程是平稳的。第四，事件随机发生——不存在使得某些子区间更可能发生的模式或聚集现象。在实际中，这些条件常常是假设的而非证实的，但验证它们是良好统计分析的一部分。
Common violations include seasonality (changing λ), dependence (e.g., contagion), or overlapping events. When data show over‑dispersion – the variance is noticeably larger than the mean – a Poisson model may be inappropriate, and a negative binomial model might be considered.
常见的违例包括季节性（λ 变化）、相依性（如传染）或事件重叠。当数据呈现过度离散——方差明显大于均值——泊松模型可能不适用，可考虑负二项分布模型。
3. Probability Mass Function | 概率质量函数
The Poisson probability mass function (PMF) gives the probability that a Poisson random variable X equals a specific count k. It is defined as:
泊松概率质量函数给出了泊松随机变量 X 等于某一特定计数值 k 的概率，定义为：
P(X = k) = (λᵏ e⁻λ) / k! , for k = 0, 1, 2, …
Here e is Euler's number (≈ 2.71828), λ is the positive rate parameter, and k! The formula combines the exponential decay e⁻λ with a power term λᵏ and normalisation by k!. The probability is always non‑negative and the sum over all k from 0 to ∞ equals 1.
这里 e 是自然常数（≈ 2.71828），λ 是正的速率参数，k! 表示 k 的阶乘。该公式结合了指数衰减因子 e⁻λ、幂项 λᵏ 以及通过 k! 进行的归一化。概率始终非负，且对所有 k 从 0 到 ∞ 的求和为 1。
. For instance, if λ = 0.5, P(X = 0) = e⁻⁰·⁵ ≈ 0.6065; if λ = 5, P(X = 5) ≈ 0.1755. Edexcel examinations expect students to compute individual probabilities using a scientific calculator or the provided cumulative Poisson tables.
随着 λ 增大，分布形状从高度偏斜（λ 较小时）逐步趋向对称（λ 大于约10时）。例如，λ = 0.5 时，P(X = 0) = e⁻⁰·⁵ ≈ 0.6065；λ = 5 时，P(X = 5) ≈ 0.1755。Edexcel 考试要求学生能够使用科学计算器或提供的泊松累积表计算各个概率值。
A short table for λ = 2.0 illustrates typical probabilities:
下表以 λ = 2.0 为例展示典型概率：
| x | P(X = x) |
|---|---|
| 0 | 0.1353 |
| 1 | 0.2707 |
| 2 | 0.2707 |
| 3 | 0.1804 |
| 4 | 0.0902 |
| 5 | 0.0361 |
4. Mean and Variance | 期望与方差
One of the most distinctive properties of the Poisson distribution is that its mean and variance are both equal to λ. For X ~ Po(λ), we have:
泊松分布最显著的性质之一是它的期望和方差相等，都等于 λ。对于 X ~ Po(λ)，有：
E(X) = λ , Var(X) = λ
This equality provides a quick diagnostic: if a sample from a supposed Poisson process yields a variance that is substantially larger than the mean (over‑dispersion), the Poisson assumption should be questioned. Conversely, if the variance is much smaller (under‑dispersion), another model may be needed.
这一等式提供了一个快速诊断方法：如果来自假定泊松过程的样本产生的方差远大于均值（过度离散），就应质疑泊松假设。反之，若方差远小于均值（离散不足），也可能需要其他模型。
the variance proportionally. For instance, if cars pass a checkpoint at an average rate of 4 per minute, then in a 5‑minute interval λ = 20, and the variance is also 20.
因此标准差为 √λ。由于 λ 是一个随区间大小伸缩的速率，改变区间会按比例改变均值和方差。例如，若汽车以平均每分钟4辆的速率通过检查点，则在5分钟区间内 λ = 20，方差也为20。
This property is essential when solving problems involving different time units: a Poisson rate per hour can be scaled down to a rate per minute simply by dividing by 60, provided the conditions still hold.
在涉及不同时间单位的问题中，这一性质至关重要：每小时泊松速率可以通过除以60简单地缩小为每分钟速率，只要条件仍然成立。
5. The Poisson Process | 泊松过程
A Poisson process is a continuous‑time counting process that satisfies the Poisson conditions. It is often used when events occur randomly in time with a constant intensity λ. The number of events in any interval of length t follows a Poisson distribution with parameter λt.
泊松过程是一个满足泊松条件的连续时间计数过程，常用于事件随时间随机发生且强度 λ 恒定的情景。任意长度为 t 的区间内的事件数量服从参数为 λt 的泊松分布。
In Edexcel questions, the concept frequently appears when students are asked to adjust λ for a new interval. For example, if the number of goals scored by a football team in 90 minutes follows Po(1.8), then the number of goals in 45 minutes is modelled as Po(0.9). The scaling must be done with care – the underlying rate per unit time remains unchanged.
在Edexcel试题中，这一概念频繁出现，通常会要求学生为新时段调整 λ。例如，某足球队在90分钟内的进球数服从 Po(1.8)，那么45分钟内的进球数可建模为 Po(0.9)。缩放时必须谨慎——单位时间的底层速率保持不变。
The waiting time until the first event in a Poisson process follows an exponential distribution, but that link is not examinable in the standard Edexcel A-Level Mathematics specification; it is nevertheless useful for conceptual understanding.
泊松过程中直到第一个事件发生的等待时间服从指数分布，但这一关联在普通Edexcel A-Level数学考纲中不作考查；不过掌握它对概念理解有益。
6. Using the Cumulative Distribution | 使用累积分布函数
babilities P(X ≤ x) or P(X < x). The cumulative distribution function (CDF) is defined as:
除了计算单个概率 P(X = k) 之外，考试题目通常还要求计算累积概率 P(X ≤ x) 或 P(X < x)。累积分布函数定义为：
P(X ≤ x) = Σk=0x (λᵏ e⁻λ) / k!
example, directly from the table for λ = 3.0, P(X ≤ 2) = 0.4232. If a problem asks for P(X ≥ 4), use the complement: P(X ≥ 4) = 1 − P(X ≤ 3).
Edexcel提供了针对一定 λ 范围（通常至10.0）的泊松累积表。学生必须能够正确查阅这些表格。例如，从 λ = 3.0 的表中可直接读出 P(X ≤ 2) = 0.4232。如果题目要求 P(X ≥ 4)，则利用补集：P(X ≥ 4) = 1 − P(X ≤ 3)。
Modern calculators (e.g. Casio ClassWiz) can also produce Poisson probabilities and cumulative probabilities directly. Nevertheless, examiners sometimes specifically instruct candidates to use tables, so familiarity with both methods is required.
现代计算器（如Casio ClassWiz）也可直接生成泊松概率和累积概率。但考官有时会明确指示考生使用统计表，因此必须同时熟悉两种方式。
7. Poisson Approximation to the Binomial | 二项分布的泊松近似
When a binomial random variable X ~ rge (typically n ≥ 50) and p small (p ≤ 0.1), with the product np not too large (often np < 10). Then, setting λ = np, we approximate X ≈ Po(np).
当二项随机变量 X ~ B(n, p) 的 n 很大、p 很小时，泊松分布可以提供方便且精确的近似。经验规则是 n 应较大（通常 n ≥ 50），p 较小（p ≤ 0.1），且乘积 np 不太大（常要求 np < 10）。然后令 λ = np，近似表示为 X ≈ Po(np)。
ile np stays constant, the binomial distribution converges to the Poisson distribution. This is extremely handy in quality control or rare‑disease studies where exact binomial calculations would be cumbersome.
其原理是：当 n → ∞、p → 0 且 np 保持不变时，二项分布收敛于泊松分布。这在质量控制或罕见病研究中极为方便，因为精确的二项计算会很繁琐。
For example, if 1% bs produced by a factory are defective, in a batch of 200 bulbs the number of defectives X ~ B(200, 0.01) can be approximated by Po(2). Then P(X = 3) ≈ (2³ e⁻²)/3! = 0.1804, close to the exact binomial value of 0.1814.
例如，某工厂生产的灯泡有1%存在缺陷，在一批200个灯泡中，缺陷数 X ~ B(200, 0.01) 可用 Po(2) 近似。则 P(X = 3) ≈ (2³ e⁻²)/3! = 0.1804，非常接近精确二项值 0.1814。
8. Sum of Independent Poisson Variables | 独立泊松变量之和
ly, then their sum Z = X + Y is also Poisson distributed with parameter λ₁ + λ₂. Formally:
如果 X 和 Y 是两个独立的泊松随机变量，参数分别为 λ₁ 和 λ₂，那么它们的和 Z = X + Y 也服从泊松分布，参数为 λ₁ + λ₂。即：
X ~ Po(λ₁), Y ~ Po(λ₂), X, Y independent → X + Y ~ Po(λ₁ + λ₂)
This additive property extends to any finite sum of independent Poisson variables. It is particularly useful when superimposing several independent Poisson processes – for instance, the total number of customers arriving at a service desk from two separate entrances, each following a Poisson stream.
这一可加性质可以推广到任意有限个独立泊松变量之和。它在叠加多个独立的泊松过程时特别有用，例如，来自两个独立入口的顾客流加总到服务台的总人数，每个入口都遵循一个泊松流。
Important caveat: independence is crucial. If X and Y are positively correlated, the variance of the sum will be larger than λ₁ + λ₂, breaking the Poisson property.
重要提醒：独立性至关重要。如果 X 和 Y 呈正相关，总和的方差会大于 λ₁ + λ₂，从而破坏泊松性质。
9. Hypothesis Testing with the Poisson Distribution | 泊松分布的假设检验
nce level. The null hypothesis typically states H₀: λ = λ₀, while the alternative can be one‑tailed (H₁: λ < λ₀ or λ > λ₀) or two‑tailed (H₁: λ ≠ λ₀).
Edexcel A-Level 统计学涵盖对泊松分布参数 λ 进行假设检验，使用预先确定的显著性水平。原假设通常为 H₀: λ = λ₀，备择假设可以是单尾的（H₁: λ < λ₀ 或 λ > λ₀）也可以是双尾的（H₁: λ ≠ λ₀）。
Once an observed count x is obtained, the p‑value or critical region is found using the Poisson distribution under H₀. For a lower‑tail test H₁: λ < λ₀, the p‑value is P(X ≤ x) under Po(λ₀). If this p‑value is less than the significance level α (commonly 0.05), H₀ is rejected. For an upper‑tail test, use P(X ≥ x) = 1 − P(X ≤ x − 1).
一旦获得观测计数 x，就可以利用 H₀ 下的泊松分布计算 p 值或确定拒绝域。对于下尾检验 H₁: λ < λ₀，p 值为 Po(λ₀) 下 P(X ≤ x)。若此 p 值小于显著性水平 α（通常为0.05），则拒绝 H₀。对于上尾检验，用 P(X ≥ x) = 1 − P(X ≤ x − 1)。
For a two‑tailed test, the p‑value is doubled (if x is on the side of lower probability), or critical regions are allocated to both tails. Students must always state the conclusion in context, e.g., 'there is sufficient evidence at the 5% level to suggest the rate has decreased.'
对于双尾检验，p 值加倍（若 x 在概率较小的一侧），或者将拒绝域分配在两个尾部。学生必须结合上下文给出结论，例如：“在5%显著性水平下，有充分证据表明该速率已下降”。
10. Modelling Assumptions and Limitations | 建模假设与局限性
Although the Poisson distribution is a powerful tool, its validity hinges on the modelling assumptions being reasonably met. In real‑life applications, events may not be perfectly independent (e.g., a car accident causing a traffic jam that triggers secondary incidents), or the rate may vary over time (rush hour vs. off‑peak).
尽管泊松分布是一个强有力的工具，但其有效性依赖于建模假设被合理满足。在现实应用中，事件可能并非完全独立（例如，一起交通事故引发交通堵塞，进而导致二次事故），或者速率可能随时间变化（高峰时段与平峰时段）。
Over‑dispersion is a common problem: when data show Var(X) > E(X), the Poisson assumption is suspect. Similarly, when events tend to arrive in clusters rather than singly, the Poisson condition of single occurrences is violated. In such cases, more advanced models like the compound Poisson or negative binomial distribution may be required.
过度离散是一个常见问题：当数据显示 Var(X) > E(X) 时，泊松假设就有疑问。类似地，当事件趋向于成群到达而非单独发生时，泊松的单发性条件就被破坏。在这些情况下，可能需要更高级的模型，如复合泊松或负二项分布。
Furthermore, the Poisson distribution is memoryless only through the exponential waiting times, but the count distribution itself is not memoryless. Checking residual plots or performing goodness‑of‑fit tests (e.g. chi‑squared test) can help assess whether a Poisson model is appropriate.
此外，泊松分布本身并不具有无记忆性（仅等待时间的指数分布具有），但计数分布本身并无此性质。通过检查残差图或进行拟合优度检验（如卡方检验）有助于评估泊松模型是否合适。
Despite these limitations, the Poisson distribution remains a cornerstone of discrete probability modelling in A‑Level Mathematics due to its simplicity and wide applicability.
尽管有这些局限性，泊松分布仍因其简洁性和广泛适用性而成为A‑Level数学中离散概率建模的基石。
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