📚 More about statistical distributions | 更多关于统计分布
Statistical distributions form the backbone of inferential statistics, allowing us to model randomness, quantify uncertainty, and draw conclusions from data. In IB Mathematics, a solid understanding of both discrete and continuous distributions, along with their properties, connections, and approximations, is essential for tackling problems in probability, hypothesis testing, and real-world applications. This article delves deeper into the behaviour, moments, and relationships of key distributions, moving beyond basic definitions to equip you with the tools needed for advanced statistical reasoning.
统计分布是推断统计学的支柱，使我们能够对随机性建模、量化不确定性并从数据中得出结论。在 IB 数学中，扎实理解离散和连续分布及其性质、关联和近似，对于解决概率、假设检验和实际应用问题至关重要。本文将深入探讨关键分布的行为、矩及相互关系，超越基本定义，为你提供进行高级统计推理所需的工具。
1. Review of Key Discrete Distributions | 关键离散分布回顾
i trials, each with success probability p. Its probability mass function (PMF) is P(X = k) = ⁿCₖ pᵏ (1−p)ⁿ⁻ᵏ, for k = 0, 1, …, n. The mean is E(X) = np and the variance is Var(X) = np(1−p).
离散型随机变量取可数个不同值。IB 课程中两种最基本的离散分布是二项分布和泊松分布。二项分布 B(n, p) 计算 n 次独立伯努利试验中的成功次数，每次成功概率为 p。其概率质量函数为 P(X = k) = ⁿCₖ pᵏ (1−p)ⁿ⁻ᵏ, k = 0, 1, …, n。均值为 E(X) = np，方差为 Var(X) = np(1−p)。
terval of time or space, given a constant average rate λ. Its PMF is P(Y = k) = e⁻λ λᵏ / k!, k = 0, 1, 2, … the mean and the variance equal λ: E(Y) = Var(Y) = λ, a property known as equidispersion.
泊松分布 Po(λ) 用于建模固定时间或空间间隔内发生的事件数，假设平均发生率为常数 λ。其概率质量函数为 P(Y = k) = e⁻λ λᵏ / k!, k = 0, 1, 2, …。值得注意的是，均值和方差都等于 λ：E(Y) = Var(Y) = λ，这一性质称为等分散性。
2. Binomial Distribution: Parameters and Properties | 二项分布：参数与性质
on arises naturally as the sum of n independent Bernoulli random variables. That is, if Xᵢ ~ Bernoulli(p) for i = 1, …, n, then X = Σ Xᵢ ~ ). This additive property extends to independent binomial variables with the same p: if X₁ ~ B(n₁, p) and X₂ ~ B(n₂, p) are independent, then X₁ + X₂ ~ B(n₁ + n₂, p).
二项分布可自然地视为 n 个独立伯努利随机变量之和。即，若 Xᵢ ~ Bernoulli(p), i = 1, …, n, 则 X = Σ Xᵢ ~ B(n, p)。这一可加性可推广到具有相同 p 的独立二项变量：若 X₁ ~ B(n₁, p) 与 X₂ ~ B(n₂, p) 独立，则 X₁ + X₂ ~ B(n₁ + n₂, p)。
The sample proportion p̂ = X / n is an unbiased estimator of p, with E(p̂) = p and Var(p̂ ften used to justify a normal approximation to the binomial, which will be explored later.
样本比例 p̂ = X / n 是 p 的无偏估计量，满足 E(p̂) = p，Var(p̂) = p(1−p)/n。当 n 较大时，二项分布的形状取决于 p：p = 0.5 时对称，否则偏斜，但偏斜度随 n 增大而减小。通常用条件 np ≥ 5 和 n(1−p) ≥ 5 来判定是否可以使用正态近似，这一点将在后文探讨。
3. Poisson Distribution and Poisson Processes | 泊松分布与泊松过程
p → 0 such that np = λ remains constant. This makes it an excellent model for rare events— ance, the number of radioactive decays per second or the number of typing errors per page.
泊松分布可以作为二项分布在 n → ∞ 且 p → 0 而 np = λ 保持常数时的极限情况推导出来。这使它成为稀有事件的绝佳模型，例如每秒放射性衰变次数或每页打字错误数。
Poisson processes have the memoryless property: the number of events in disjoint intervals are independent, and the distribution depends only on the length of the interval. If events occur at a rate λ per unit time, then the number in an interval of length t follows Po(λt). Furthermore, the sum of independent Poisson variables is also Poisson: if Y₁ ~ Po(λ₁) and Y₂ ~ Po(λ₂) are independent, then Y₁ + Y₂ ~ Po(λ₁ + λ₂).
泊松过程具有无记忆性：不相交区间内的事件数相互独立，且分布仅取决于区间长度。若事件以每单位时间 λ 的速率发生，则在长度为 t 的区间内的事件数服从 Po(λt)。此外，独立泊松变量之和仍为泊松变量：若 Y₁ ~ Po(λ₁) 与 Y₂ ~ Po(λ₂) 独立，则 Y₁ + Y₂ ~ Po(λ₁ + λ₂)。
Because λ equals both the mean and variance, a data set with sample variance significantly larger than its mean may indicate overdispersion, meaning a Poisson model is inappropriate. This diagnostic check is fundamental when applying the distribution to real data.
由于 λ 同时是均值和方差，若数据集的样本方差远大于其均值，则可能表明存在过度分散，意味着泊松模型不适用。这一诊断性检查在将分布应用于实际数据时至关重要。
4. The Normal Distribution: Properties and Standardisation | 正态分布：性质与标准化
ole in the Central Limit Theorem. Its probability density function (PDF) is f(x) = (1 / (σ√2π)) e−½ ((x−μ)/σ)² e is bell-shaped, symmetric about the mean μ, and the points of inflection occur at μ ± σ.
正态分布 N(μ, σ²) 是统计学中最重要的连续分布，因为它在中心极限定理中扮演核心角色。其概率密度函数为 f(x) = (1 / (σ√2π)) e−½ ((x−μ)/σ)²。曲线呈钟型，关于均值 μ 对称，拐点位于 μ ± σ 处。
ndardised using the transformation Z = (X − μ) / σ, producing a standard normal variable Z ~ mal tables to find probabilities and percentiles. Key properties include: approximately 68% of values lie within μ ± σ, 95% within μ ± 2σ, and 99.7% within μ ± 3σ (the empirical rule).
任何正态变量都可以通过变换 Z = (X − μ) / σ 进行标准化，得到标准正态变量 Z ~ N(0, 1)。标准化使我们能够使用标准正态表查找概率和百分位数。关键性质包括：约 68% 的值落在 μ ± σ 内，95% 落在 μ ± 2σ 内，99.7% 落在 μ ± 3σ 内（经验法则）。
Linear combinations of independent normal variables are also normally distributed. If X₁ ~ N(μ₁, σ₁²) and X₂ ~ N(μ₂, σ₂²) are independent, then aX₁ + bX₂ + c ~ N(aμ₁ + bμ₂ + c, a²σ₁² + b²σ₂²). This property is extremely useful in constructing confidence intervals and test statistics.
独立正态变量的线性组合仍服从正态分布。若 X₁ ~ N(μ₁, σ₁²) 与 X₂ ~ N(μ₂, σ₂²) 独立，则 aX₁ + bX₂ + c ~ N(aμ₁ + bμ₂ + c, a²σ₁² + b²σ₂²)。这一性质在构造置信区间和检验统计量时极为有用。
5. Central Limit Theorem | 中心极限定理
ulation with mean μ and finite variance σ², the sampling distribution of the sample mean X̄ rposes, n ≥ 30 is considered sufficiently large, but if the population is itself normal, X̄ is exactly normal for any n.
中心极限定理（CLT）指出，对于从均值为 μ、方差为 σ²（有限）的任意总体中抽取的容量为 n 的随机样本，样本均值 X̄ 的抽样分布随着 n 增大而趋近于正态分布 N(μ, σ²/n)，无论原始总体形状如何。在大多数实际应用中，n ≥ 30 被认为是足够大的，但如果总体本身是正态的，则 X̄ 对任何 n 都精确服从正态分布。
The CLT justifies the widespread use of normal-based procedures in statistics even when the underlying data are non-normal. For example, in quality control, the mean of a large number of measurements can be modelled by a normal distribution, enabling the construction of control charts. The theorem also underpins the normal approximation to the binomial and Poisson distributions for large sample sizes or means.
中心极限定理为即使底层数据非正态时也在统计中广泛使用基于正态的方法提供了依据。例如，在质量控制中，大量测量值的均值可以用正态分布建模，从而构建控制图。该定理也是大样本或大均值下二项分布和泊松分布正态近似的基础。
6. Sampling Distributions: Sample Mean and Proportion | 抽样分布：样本均值与样本比例
distribution is called the sampling distribution of that statistic. For the sample mean X̄, we have E(X̄) = μ and Var(X̄) = σ²/n, so the standard error is σ/√n. If the population is normal, X̄ is normal; otherwise, the CLT guarantees approximate normality for n ≥ 30.
当我们从总体中反复抽取样本并计算一个统计量时，所得到的概率分布称为该统计量的抽样分布。对于样本均值 X̄，有 E(X̄) = μ，Var(X̄) = σ²/n，因此标准误差为 σ/√n。若总体为正态，则 X̄ 服从正态分布；否则，中心极限定理保证在 n ≥ 30 时近似正态。
The sample proportion p̂ riance p(1−p)/n. When the sample size is large enough (typically np ≥ 5 and n(1−p) ≥ 5), p̂ is approximately normal: p̂ ~ cial for constructing confidence intervals and performing hypothesis tests on proportions.
样本比例 p̂ 是二进制数据样本均值的特例。其抽样分布的均值为 p，方差为 p(1−p)/n。当样本量足够大时（通常要求 np ≥ 5 且 n(1−p) ≥ 5），p̂ 近似服从正态分布：p̂ ~ N(p, p(1−p)/n)。这一结果对于构造比例的置信区间和进行假设检验至关重要。
7. Chi-Squared and Student's t Distributions | 卡方分布与学生氏 t 分布
The chi-squared (χ²) distribution arises naturally in inference. If Z₁, Z₂, …, Zν are independent standard normal variables, then the sum of their squares follows a chi-squared distribution with ν degrees of freedom: Σ Zᵢ² ~ χ²(ν). The mean of a χ²(ν) distribution is ν and the variance is 2ν. The distribution is skewed to the right and becomes more symmetric as ν increases.
卡方（χ²）分布在统计推断中自然产生。若 Z₁, Z₂, …, Zν 是独立的标准正态变量，则它们的平方和服从自由度为 ν 的卡方分布：Σ Zᵢ² ~ χ²(ν)。χ²(ν) 分布的均值为 ν，方差为 2ν。该分布右偏，且随着 ν 增大而更趋对称。
Student' are root of an independent chi-squared variable divided by its degrees of freedom: T = Z / √U/ν, where Z ~ N(0,1), U ~ sed to construct confidence intervals and tests for the population mean when σ is unknown.
学生氏 t 分布定义为标准正态变量与独立卡方变量除以其自由度后的平方根之比：T = Z / √U/ν，其中 Z ~ N(0,1)，U ~ χ²(ν)，且 Z 与 U 独立。t 分布的尾部比正态分布更厚，但随着 ν → ∞ 趋近于 N(0,1)。当 σ 未知时，它被用来构造总体均值的置信区间和假设检验。
The sample variance (n−1)S² / σ² from a normal population follows a χ² distribution with n−1 degrees of freedom. This fact is the foundation of confidence intervals for σ² and of the χ² test for independence and goodness-of-fit.
来自正态总体的样本方差 (n−1)S² / σ² 服从自由度为 n−1 的卡方分布。这一事实是 σ² 置信区间以及独立性检验与拟合优度检验的卡方检验的基础。
8. Distribution of Sums and Linear Combinations | 和与线性组合的分布
A fundamental result is that the sum of independent normal random variables is normal: if Xᵢ ~ N(μᵢ, σᵢ²) are independent, then Σ Xᵢ ~ N(Σ μᵢ, Σ σᵢ²). This extends to any linear combination: Σ aᵢ Xᵢ + b ~ N(Σ aᵢ μᵢ + b, Σ aᵢ² σᵢ²). This property, together with the CLT, makes linear combinations the workhorse of parametric statistics.
一个基本结果是，独立正态随机变量之和服从正态分布：若 Xᵢ ~ N(μᵢ, σᵢ²) 独立，则 Σ Xᵢ ~ N(Σ μᵢ, Σ σᵢ²)。这可以推广到任何线性组合：Σ aᵢ Xᵢ + b ~ N(Σ aᵢ μᵢ + b, Σ aᵢ² σᵢ²)。这一性质与中心极限定理一起，使线性组合成为参数统计的主力工具。
For discrete distributions, the sum of independent Poisson variables is Poisson (Po(λ₁) + Po(λ₂) ~ Po(λ₁+λ₂)), and the sum of independent binomial variables with the same p is binomial. However, the sum of binomial variables with different p does not follow a simple closed-form distribution, which has implications in meta-analysis and modelling heterogeneity.
对于离散分布，独立泊松变量之和为泊松分布（Po(λ₁) + Po(λ₂) ~ Po(λ₁+λ₂)），具有相同 p 的独立二项变量之和为二项分布。然而，具有不同 p 的二项变量之和并没有简单的闭式分布，这在荟萃分析和异质性建模中具有重要影响。
When variables are not independent, the variance of their sum includes covariance terms: Var(X̄) for dependent data must account for intra-class correlation. In IB-level problems, independence is usually assumed, but understanding when this assumption fails is an advanced statistical skill.
当变量不独立时，其和的方差包含协方差项：非独立数据的 Var(X̄) 必须考虑组内相关性。在 IB 级别的问题中通常假设独立，但理解该假设何时失效是一项高级统计技能。
9. Approximations Between Distributions | 分布之间的近似
p ≤ 0.05, or n ≥ 100 and np ≤ 10. This is useful for rare event probability calculations.
几种分布近似可以简化计算，并在假设检验中必不可少。二项分布的泊松近似适用于 n 很大而 p 很小的情况，令 λ = np。常用的经验法则是 n ≥ 20 且 p ≤ 0.05，或 n ≥ 100 且 np ≤ 10。这种近似对稀有事件概率计算很有用。
valid when both np and n(1−p) are at least 5 (or more conservatively 10). In that case, X ~ roximated by a continuous one. For example, P(X ≤ k) ≈ P(Z ≤ (k + 0.5 − np) / √(np(1−p))).
二项分布的正态近似在 np 和 n(1−p) 均至少为 5（更保守地 10）时适用。此时 X ~ B(n, p) 近似为 N(np, np(1−p))。由于用连续分布近似离散分布，需通过将边界调整 0.5 来进行连续性校正。例如，P(X ≤ k) ≈ P(Z ≤ (k + 0.5 − np) / √(np(1−p)))。
(k + 0.5 − λ) / √λ). These approximations become more accurate as the parameters increase.
类似地，当 λ 较大（如 λ > 10）时，泊松分布可用 N(λ, λ) 近似，同样应用连续性校正：P(Y ≤ k) ≈ P(Z ≤ (k + 0.5 − λ) / √λ)。随着参数增大，这些近似变得更加精确。
10. Applications in Hypothesis Testing | 假设检验中的应用
opulation mean uses the distribution of the sample mean: under the null hypothesis, Z = (X̄ − μ₀) / (σ/√n) ~ σ is unknown, the t-test replaces σ with the sample standard deviation s, yielding T = (X̄ − μ₀) / (s/√n) ~ t(n−1). The p-value is computed as the tail probability of the corresponding distribution.
统计分布并非孤立学习，它们是推断检验的引擎。总体均值的单样本 z 检验使用样本均值的分布：在原假设下，Z = (X̄ − μ₀) / (σ/√n) ~ N(0,1)。当 σ 未知时，t 检验用样本标准差 s 代替 σ，得到 T = (X̄ − μ₀) / (s/√n) ~ t(n−1)。p 值即为相应分布的尾部概率。
Oᵢ − Eᵢ)² / Eᵢ, under the null hypothesis, approximately follows a χ² distribution with appropriate degrees of freedom. Large values indicate a poor fit, and the critical region is determined from the tail of the χ² distribution.
独立性或拟合优度的卡方检验依赖 χ² 分布。检验统计量 χ² = Σ (Oᵢ − Eᵢ)² / Eᵢ 在原假设下近似服从具有适当自由度的卡方分布。大的值表明拟合不佳，拒绝域由 χ² 分布的尾部确定。
Understanding which distribution to use, when approximations are valid, and how to interpret tail areas is the mark of a proficient statistics student. Mastery of these topics not only ensures success in examinations but also provides a solid foundation for further study in data science, economics, and the natural sciences.
理解何时使用哪种分布、近似何时有效以及如何解释尾部区域，是一个熟练统计学子的标志。掌握这些主题不仅能确保考试成功，也为数据科学、经济学和自然科学领域的进一步学习奠定了坚实基础。
Published by TutorHao | IB Statistics 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