📚 Hypothesis Testing for CIE A-Level Mathematics | A-Level CIE 数学：假设检验考点精讲
Hypothesis testing is a cornerstone of statistical inference and a major topic in the CIE A-Level Mathematics (9709) syllabus. It provides a formal framework for making decisions about population parameters based on sample data. In Papers 5 (S1) and 6 (S2), students must master tests for proportions (binomial), normal means, Poisson rates, t‑tests, and chi‑squared methods. This article distills the essential concepts, procedures, and common pitfalls, equipping you with a clear revision guide for all hypothesis‑testing scenarios you will encounter.
假设检验是统计推断的核心，也是 CIE A-Level 数学 (9709) 的重中之重。它提供了一个基于样本数据对总体参数做出决策的正式框架。在试卷5 (S1) 和试卷6 (S2) 中，同学们需要熟练掌握比例（二项分布）检验、正态均值检验、泊松分布检验、t 检验以及卡方检验。本文精炼了关键概念、标准步骤和常见易错点，为你梳理出一份涵盖所有假设检验考点的清晰复习指南。
1. Fundamentals of Hypothesis Testing | 假设检验基础
A hypothesis test begins with two competing statements: the null hypothesis H₀, which assumes no effect or no difference, and the alternative hypothesis H₁, which represents what we seek evidence for. We collect sample data and assess how likely the observed result (or something more extreme) would be if H₀ were true. If this probability is very small, we reject H₀ in favour of H₁.
假设检验始于两个对立的陈述：原假设 H₀，通常表示没有效应或没有差异；备择假设 H₁，表示我们希望找到证据支持的主张。我们收集样本数据，然后评估在原假设为真的条件下，观察到当前结果（或更极端结果）的概率。如果这个概率非常小，我们就拒绝原假设而接受备择假设。
The decision is always made in the context of a pre‑chosen significance level α, often 5% or 1%. The test is not designed to “prove” H₁; it merely assesses whether the data provide sufficient evidence against H₀.
决策总是在事先选定的显著性水平 α 下进行，常见取值为 5% 或 1%。检验并非为了“证明” H₁ 成立，而是判断数据是否提供了足够的证据来否定 H₀。
2. One-tailed and Two-tailed Tests | 单尾与双尾检验
The form of H₁ determines whether the test is one‑tailed or two‑tailed. A one‑tailed test investigates a specific direction, such as H₁: p > 0.5 or H₁: μ < 100. A two‑tailed test simply detects any difference, e.g., H₁: p ≠ 0.5 or H₁: μ ≠ 100. The choice affects how the rejection region and p‑value are constructed.
备择假设的形式决定了检验是单尾还是双尾。单尾检验考察特定的方向，例如 H₁: p > 0.5 或 H₁: μ < 100。双尾检验仅检测是否存在差异，例如 H₁: p ≠ 0.5 或 H₁: μ ≠ 100。这一选择会影响拒绝域和 p 值的构建方式。
In a two‑tailed test at significance level α, the total rejection region is split equally between the two tails, so each tail covers an area of α/2. In contrast, a one‑tailed test places the entire α in the relevant tail. Consequently, a one‑tailed test is more “power‑efficient” if the direction is correctly specified, but it must be justified by the context of the problem.
在显著性水平为 α 的双尾检验中，总的拒绝域被平分为两个尾部，每个尾部覆盖 α/2 的面积。相反，单尾检验将整个 α 置于感兴趣的那个尾部。因此，如果方向规定正确，单尾检验的“功效”更高，但必须能依据问题的背景给出合理解释。
3. Significance Level, Critical Region and Critical Value | 显著性水平、拒绝域与临界值
The significance level α is the maximum probability we are willing to accept for making a Type I error (rejecting a true H₀). The critical region (or rejection region) is the set of values of the test statistic that leads to the rejection of H₀. The boundary of this region is the critical value. If the test statistic falls inside the critical region, we reject H₀; otherwise, we do not reject H₀.
显著性水平 α 是我们愿意承担的第一类错误（弃真）的最大概率。拒绝域（临界域）是检验统计量的取值集合，当统计量落入该区域时就拒绝 H₀。该区域的边界就是临界值。如果检验统计量落在拒绝域内，我们就拒绝 H₀；否则不拒绝 H₀。
For example, in a one‑tailed binomial test with H₁: p > 0.5, n = 20 and α = 0.05, we find the smallest integer c such that P(X ≥ c | p = 0.5) ≤ 0.05. That c becomes the critical value, and the critical region is X ≥ c. The exact probability of X ≥ c is the actual significance level of the test.
例如，在一个单尾二项检验中，H₁: p > 0.5，n = 20，α = 0.05，我们需要找到最小的整数 c，使得 P(X ≥ c | p = 0.5) ≤ 0.05。该 c 就是临界值，拒绝域为 X ≥ c。而 P(X ≥ c) 的精确概率就是检验的实际显著性水平。
4. P-value Method | P 值法
The p‑value is the probability, under the assumption that H₀ is true, of obtaining a result at least as extreme as the observed test statistic. Instead of comparing the test statistic with a critical value, we compare the p‑value directly with α: if p ≤ α, we reject H₀; if p > α, we do not reject H₀. The p‑value approach is extremely popular because it gives a measure of the strength of evidence against H₀.
p 值是指在原假设 H₀ 为真的前提下，获得与观测到的检验统计量同样极端或更极端结果的概率。我们无需再去与临界值比较，而是直接将 p 值与 α 比较：若 p ≤ α，则拒绝 H₀；若 p > α，则不拒绝 H₀。p 值法的优势在于它直接量化了反对 H₀ 的证据强度。
In a two‑tailed test, the p‑value is the probability in both tails: if the test statistic is symmetric, it is twice the probability in the observed tail. For example, in a normal test with H₁: μ ≠ μ₀, if the observed z‑score is 2.10, p‑value = 2 × P(Z > 2.10). CIE exam questions frequently ask you to calculate and interpret the p‑value.
在双尾检验中，p 值为两个尾部概率之和：若检验统计量对称分布，则为观测值所在尾部概率的两倍。例如，在 H₁: μ ≠ μ₀ 的正态检验中，若观察到 z = 2.10，则 p 值 = 2 × P(Z > 2.10)。CIE 试题经常要求计算并解释 p 值。
5. Type I and Type II Errors | 第一类错误与第二类错误
- Type I error: Rejecting H₀ when it is actually true. The probability is α, the significance level.
- Type II error: Failing to reject H₀ when H₁ is true. The probability is denoted β.
- Power of a test is 1 − β, the probability of correctly rejecting a false H₀.
- 第一类错误：原假设为真却被拒绝，概率为 α，即显著性水平。
- 第二类错误：备择假设为真却未能拒绝原假设，概率记为 β。
- 检验的功效 (power) 为 1 − β，即正确拒绝错误原假设的概率。
Minimising α and β simultaneously is impossible for a fixed sample size; reducing α increases β. Increasing the sample size is the best way to reduce both. CIE S2 questions often ask you to calculate the probability of a Type II error for a specific alternative, or to find the critical region that satisfies a given α while illustrating the effect on β.
对于固定的样本量，同时最小化 α 和 β 是不可能的；减小 α 会增大 β。增加样本量是同时降低两者的最佳途径。CIE S2 的考题常要求针对特定的备择假设计算第二类错误的概率，或在满足给定 α 的前提下确定拒绝域，同时说明对 β 的影响。
6. Binomial Hypothesis Testing | 二项分布假设检验
When testing a population proportion p, we use the binomial distribution X ~ B(n, p). The null hypothesis typically states H₀: p = p₀. The alternative can be one‑tailed (p > p₀ or p < p₀) or two‑tailed (p ≠ p₀). Because the binomial distribution is discrete, the desired significance level α is rarely achieved exactly; we use the largest critical region such that the probability of Type I error is ≤ α.
检验总体比例 p 时，我们使用二项分布 X ~ B(n, p)。原假设通常为 H₀: p = p₀。备择假设可以是单尾 (p > p₀ 或 p < p₀) 或双尾 (p ≠ p₀)。由于二项分布是离散的，精确达到名义显著性水平 α 的情况很少见；我们通常取使第一类错误概率不超过 α 的最大拒绝域。
To construct the critical region for H₁: p > p₀, find the smallest integer r such that P(X ≥ r | p = p₀) ≤ α. The critical region is X ≥ r, and the critical value is r. For H₁: p < p₀, use the largest integer s such that P(X ≤ s) ≤ α. For a two‑tailed test, allocate α/2 to each tail. The exam may also ask you to find the p‑value by calculating the probability of the observed or more extreme outcomes.
若要构建 H₁: p > p₀ 的拒绝域，找出最小的整数 r 使得 P(X ≥ r | p = p₀) ≤ α。拒绝域为 X ≥ r，临界值即为 r。对于 H₁: p < p₀，找出最大的整数 s 使得 P(X ≤ s) ≤ α。对于双尾检验，每个尾部分配 α/2。考试也可能要求计算 p 值，即计算观测值以及更极端结果的概率。
Example: n = 25, H₀: p = 0.4, H₁: p > 0.4, α = 0.05. P(X ≥ 14) = 1 − P(X ≤ 13) ≈ 0.043. Since 0.043 ≤ 0.05, critical region is X ≥ 14. If we observe x = 15, we reject H₀.
示例： n = 25，H₀: p = 0.4，H₁: p > 0.4，α = 0.05。P(X ≥ 14) = 1 − P(X ≤ 13) ≈ 0.043。因为 0.043 ≤ 0.05，拒绝域为 X ≥ 14。若观察到 x = 15，则拒绝 H₀。
7. Normal Distribution Hypothesis Test for a Mean (Variance Known) | 正态分布均值检验（方差已知）
When the population is normal with known variance σ², or when the sample is large enough for the Central Limit Theorem to apply, we test the mean μ using the z‑test. The test statistic is
z = (x̄ − μ₀) / (σ/√n)
where x̄ is the sample mean, μ₀ is the hypothesised mean, σ is the population standard deviation, and n is the sample size. Under H₀, Z ~ N(0, 1).
当总体服从正态分布且方差 σ² 已知，或样本量足够大以至于中心极限定理适用时，我们使用 z 检验来检验均值 μ。检验统计量为
z = (x̄ − μ₀) / (σ/√n)
其中 x̄ 是样本均值，μ₀ 是假设的总体均值，σ 是总体标准差，n 是样本容量。在原假设下，Z ~ N(0, 1)。
For a one‑tailed test, compare z with the standard normal critical value z_α (e.g., 1.645 for α = 0.05 upper‑tail). For a two‑tailed test, use z_{α/2} (e.g., ±1.96). Reject H₀ if z falls in the critical region. The p‑value is obtained from Φ(z) or 2×(1 − Φ(|z|)) for two‑sided tests. CIE S1 tests often involve finding the probability of sample means falling in a given range.
对于单尾检验，将 z 与标准正态临界值 z_α 比较（如 α = 0.05 上尾检验，z_α = 1.645）。双尾检验则使用 z_{α/2}（如 ±1.96）。若 z 落入拒绝域则拒绝 H₀。p 值可通过标准正态分布表得到：单尾为 Φ(z) 或 1 − Φ(z)，双尾为 2×(1 − Φ(|z|))。CIE S1 考试中常出现求样本均值落在某范围的概率问题。
8. t-test for a Population Mean (Variance Unknown) | 总体均值的 t 检验（方差未知）
When the population variance is unknown and must be estimated by the sample variance s², the test statistic follows a t‑distribution with ν = n − 1 degrees of freedom. The statistic is
t = (x̄ − μ₀) / (s/√n)
This t‑test assumes that the underlying population is normally distributed. The t‑distribution is wider than the normal distribution, reflecting the additional uncertainty from estimating σ.
当总体方差未知，需要用样本方差 s² 估计时，检验统计量服从自由度为 ν = n − 1 的 t 分布。统计量为
t = (x̄ − μ₀) / (s/√n)
该 t 检验要求总体近似服从正态分布。t 分布比正态分布更“宽”，体现了由于估计 σ 带来的额外不确定性。
Critical values are obtained from t‑tables using ν. For example, with n = 10, ν = 9, a two‑tailed α = 0.05 gives critical t ≈ ±2.262. If the computed |t| exceeds the critical value, we reject H₀. The p‑value approach requires a calculator or statistical tables. CIE S2 often embeds this test in contextual problems where variance is unknown, and you must state assumptions clearly.
临界值通过自由度 ν 并查 t 分布表获得。例如 n = 10, ν = 9，双尾 α = 0.05 对应的临界 t 值约为 ±2.262。若计算出的 |t| 大于临界值，则拒绝 H₀。p 值法通常需要计算器或详细统计表。CIE S2 经常在方差未知的实际背景下出现此类检验，并要求你明确陈述假定条件。
9. Poisson Hypothesis Testing | 泊松分布假设检验
For count data that can be modelled by a Poisson distribution Po(λ), we test the rate parameter λ. The null is H₀: λ = λ₀. Given n observations or a total count T, we can sum events; the total T ~ Po(nλ₀) under H₀. The critical region is built using cumulative Poisson probabilities, analogous to the binomial case, but now for a mean rate.
对于能用泊松分布 Po(λ) 建模的计数数据，我们对速率参数 λ 进行检验。原假设为 H₀: λ = λ₀。若获得 n 次观测或总计数 T，在 H₀ 下总和 T ~ Po(nλ₀)。利用累积泊松概率构建拒绝域，方法与二项情形类似，但此时针对的是平均发生率。
Example: A machine produces flaws at a rate of 0.5 per metre. In a 20‑metre length, 15 flaws are found. Test if the rate has increased (α = 0.05). H₀: λ = 0.5, H₁: λ > 0.5. Under H₀, T ~ Po(20×0.5 = 10). P(T ≥ 15) = 1 − P(T ≤ 14). Using tables, P(T ≤ 14) ≈ 0.9166, so p‑value ≈ 0.0834 > 0.05, do not reject H₀.
示例：一台机器平均每米出现 0.5 个瑕疵。在 20 米布上发现了 15 个瑕疵。检验瑕疵率是否上升 (α = 0.05)。H₀: λ = 0.5，H₁: λ > 0.5。在 H₀ 下，T ~ Po(20×0.5 = 10)。P(T ≥ 15) = 1 − P(T ≤ 14)。查表得 P(T ≤ 14) ≈ 0.9166，p 值 ≈ 0.0834 > 0.05，不拒绝 H₀。
In two‑tailed Poisson tests, symmetry cannot be assumed, so find the largest lower tail with probability ≤ α/2 and the largest upper tail with probability ≤ α/2.
在双尾泊松检验中，不能假定对称性，因此需取累积概率 ≤ α/2 的最大下尾区域和 ≤ α/2 的最大上尾区域。
10. Chi-squared Goodness-of-Fit Test | 卡方拟合优度检验
The χ² goodness‑of‑fit test checks whether an observed frequency distribution matches a theoretical model. The test statistic is
χ² = ∑ (Oᵢ − Eᵢ)² / Eᵢ
where Oᵢ are observed frequencies and Eᵢ are expected frequencies under H₀. The statistic approximately follows a χ² distribution with ν = k − 1 − m degrees of freedom, where k is the number of categories and m is the number of estimated parameters.
卡方拟合优度检验用于检验观测频数分布是否符合某个理论模型。检验统计量为
χ² = ∑ (Oᵢ − Eᵢ)² / Eᵢ
其中 Oᵢ 为观测频数，Eᵢ 为 H₀ 下的期望频数。该统计量近似服从 χ² 分布，自由度 ν = k − 1 − m，k 为类别数，m 为需要估计的参数个数。
Crucially, all expected frequencies should be at least 5. If any Eᵢ < 5, categories must be combined. The null is rejected for large values of χ² (upper‑tail test only). The critical value χ²_ν(α) is found from tables. CIE exam questions regularly provide the observed data and ask you to calculate expected values, the test statistic, degrees of freedom, and conclusion. Sometimes you need to test a specific distribution, like uniform or binomial.
关键要求是，所有期望频数应至少为 5；若某个 Eᵢ < 5，则必须合并类别。原假设在 χ² 值过大时被拒绝（仅单尾检验）。临界值 χ²_ν(α) 由卡方分布表查得。CIE 试题通常给出观测数据，要求你计算期望频数、检验统计量、自由度并得出结论。有时还需检验特定分布，如均匀分布或二项分布。
11. Chi-squared Test for Independence (Contingency Tables) | 卡方独立性检验（列联表）
This test investigates whether two categorical variables are independent. Data are arranged in an r × c contingency table. Expected frequencies are computed from row and column totals:
Eᵢⱼ = (row total × column total) / grand total
The same χ² formula applies. The degrees of freedom are ν = (r − 1)(c − 1). Again, no expected value should fall below 5.
该检验考察两个分类变量是否独立。数据以 r × c 列联表呈现。期望频数由行列合计数推算：
Eᵢⱼ = (行合计 × 列合计) / 总合计
使用相同的 χ² 公式，自由度为 ν = (r − 1)(c − 1)。同样，期望频数不得低于 5。
If the calculated χ² exceeds the critical value, we reject the null hypothesis of independence and conclude that an association exists. CIE often sets problems that require you to fill in the contingency table, calculate expected frequencies, and then perform the test. Pay attention to rounding and the statement of conclusions in the context of the problem. A common pitfall is confusing independence with a causal relationship – the test only reveals association, not cause.
若计算得到的 χ² 超过临界值，便拒绝独立性原假设，并得出存在关联的结论。CIE 常设置需要补全列联表、计算期望频数再进行检验的题目。注意舍入，并根据问题上下文陈述结论。一个常见误区是将独立性误认为因果关系——检验仅揭示关联，而非因果。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导Cancel reply