Hypothesis Testing | 假设检验

📚 Hypothesis Testing | 假设检验

Hypothesis testing is a fundamental statistical method used to decide whether there is enough evidence from a sample to reject a claim about a population parameter. It forms a core part of the Edexcel A-Level Mathematics syllabus, enabling students to make informed conclusions based on data. From drug trials to manufacturing quality control, hypothesis testing underpins scientific reasoning and data-driven decisions. Mastering this topic involves understanding null and alternative hypotheses, significance levels, probability calculations, and the interpretation of results in context.

假设检验是一种基本的统计方法，用于判断样本是否有足够证据拒绝关于总体参数的某个主张。它是Edexcel A-Level数学大纲的核心内容，使学生能够基于数据做出有根据的结论。从药物试验到制造业质量控制，假设检验是科学推理与数据驱动决策的基础。掌握这一主题需要理解原假设与备择假设、显著性水平、概率计算以及在上下文中解释结果。

---

1. The Core Idea of Hypothesis Testing | 假设检验的核心思想

A hypothesis test begins with a statement about a population parameter, such as a proportion p or a mean μ. We collect sample data and calculate a test statistic. If the sample result is very unlikely to occur purely by chance under the initial assumption, we reject that assumption and conclude that something else is happening. The logic mirrors a court trial: the defendant is assumed innocent until the evidence proves guilt beyond reasonable doubt.

假设检验从一个关于总体参数的陈述开始，例如比例 p 或均值 μ。我们收集样本数据并计算检验统计量。如果样本结果在初始假设下纯粹由偶然性发生的概率非常小，我们就拒绝该假设，认为存在其他情况。这种逻辑类似于法庭审判：被告被假定无罪，直到证据在合理怀疑之外证明有罪。

---

2. Null and Alternative Hypotheses | 原假设与备择假设

Every hypothesis test has two competing statements. The null hypothesis, denoted H₀, is the conservative or status-quo claim that we assume to be true. The alternative hypothesis, denoted H₁ or Hₐ, represents what we suspect might be true instead. For example, when testing whether a coin is biased, we might set H₀: p = 0.5 and H₁: p ≠ 0.5. The aim of the test is to decide whether the data provide sufficient evidence to reject H₀ in favour of H₁.

每个假设检验都有两个相互竞争的陈述。原假设，记为 H₀，是我们假定为真的保守或现状主张。备择假设，记为 H₁ 或 Hₐ，代表我们怀疑可能成立的替代主张。例如，在检验一枚硬币是否偏斜时，我们可以设 H₀: p = 0.5 和 H₁: p ≠ 0.5。检验的目的是判断数据是否提供足够的证据拒绝 H₀ 而支持 H₁。

---

3. Significance Level and Critical Region | 显著性水平与临界区域

The significance level, denoted α (alpha), is the probability of rejecting H₀ when it is actually true. Common choices are 0.05 (5%) or 0.01 (1%). Before conducting a test, we determine α and define a critical region — the set of values of the test statistic that leads to the rejection of H₀. If the test statistic falls inside the critical region, we reject H₀; otherwise, we do not reject it. The boundary of this region is called the critical value.

显著性水平，记为 α（alpha），是当 H₀ 实际为真时拒绝 H₀ 的概率。常用的选择有 0.05（5%）或 0.01（1%）。在进行检验之前，我们确定 α 并定义临界区域——即导致拒绝 H₀ 的检验统计量取值集合。如果检验统计量落在临界区域内，我们拒绝 H₀；否则，我们不拒绝 H₀。该区域的边界称为临界值。

---

4. One-Tailed and Two-Tailed Tests | 单尾检验与双尾检验

When the alternative hypothesis specifies a direction (e.g., H₁: p > 0.5), we use a one-tailed test. The entire significance level α is placed in one tail of the distribution. When the alternative hypothesis is non-directional (e.g., H₁: p ≠ 0.5), a two-tailed test is used, splitting α equally between both tails. Edexcel questions often require students to choose the appropriate test based on the wording of the problem, such as ‘increase’, ‘improve’ implying one-tailed, while ‘change’, ‘different’ imply two-tailed.

当备择假设指明方向（例如 H₁: p > 0.5）时，我们使用单尾检验。整个显著性水平 α 被放置在分布的一个尾部。当备择假设无方向性（例如 H₁: p ≠ 0.5）时，使用双尾检验，将 α 均分至两个尾部。Edexcel 试题常要求学生根据问题的措辞选择合适的检验方式，如“增加”、“提高”暗示单尾，而“改变”、“不同”则暗示双尾。

---

5. Test Statistic and P-Value | 检验统计量与 P 值

The test statistic is a value calculated from the sample data, used to decide the outcome of the test. In binomial tests, it is often the number of successes X. In normal tests, it may be the sample mean X̄ or the standardized z-score. The p-value is the probability of observing a test statistic at least as extreme as the one obtained, assuming H₀ is true. A small p-value (less than α) indicates that such an extreme result is unlikely under H₀, leading to rejection.

检验统计量是从样本数据计算得出的值，用于决定检验的结果。在二项检验中，它通常是成功次数 X。在正态检验中，它可能是样本均值 X̄ 或标准化的 z 分数。P 值是假定 H₀ 为真时，观察到至少与当前结果同样极端的检验统计量的概率。较小的 p 值（小于 α）表明在 H₀ 下出现如此极端的结果不太可能，从而得出拒绝 H₀ 的结论。

---

6. Errors in Hypothesis Testing | 假设检验中的错误

Two types of errors can occur. A Type I error happens when we reject H₀ when it is actually true; its probability is exactly α. A Type II error occurs when we fail to reject H₀ when H₁ is true; its probability is denoted β (beta). The power of a test is 1 − β, which is the probability of correctly rejecting a false H₀. In A-Level questions, you may be asked to explain these errors in context, such as convicting an innocent person (Type I) or letting a guilty person go free (Type II).

可能发生两种错误。第一类错误发生在 H₀ 为真时我们却拒绝了它；其概率恰好为 α。第二类错误发生在 H₁ 为真时我们未能拒绝 H₀；其概率记为 β（beta）。检验的功效为 1 − β，即正确拒绝错误 H₀ 的概率。在 A-Level 考题中，你可能需要结合上下文解释这些错误，例如将无辜者定罪（第一类错误）或让有罪者逍遥法外（第二类错误）。

---

7. Hypothesis Testing for a Binomial Proportion | 二项分布比例的假设检验

When testing a population proportion, the number of successes X out of n trials follows a binomial distribution B(n, p). We compute the probability of obtaining the observed value or a more extreme value under the null hypothesis H₀: p = p₀. For a one-tailed test, we sum probabilities in the direction specified. For example, if H₁: p < 0.3 with n=20 and we observed X=3, we calculate P(X ≤ 3 | p=0.3). If this probability is less than α, we reject H₀. Edexcel expects careful use of the binomial cumulative distribution tables or calculators.

在检验总体比例时，n 次试验中的成功次数 X 服从二项分布 B(n, p)。我们在原假设 H₀: p = p₀ 下计算获得观察值或更极端值的概率。对于单尾检验，我们按指定方向对概率求和。例如，若 H₁: p < 0.3，n=20 且观察到 X=3，我们计算 P(X ≤ 3 | p=0.3)。如果该概率小于 α，则拒绝 H₀。Edexcel 要求谨慎使用二项累积分布表或计算器。

---

8. Hypothesis Testing for a Normal Mean (Known Variance) | 正态均值的假设检验（已知方差）

When the population is normally distributed and the variance σ² is known, the sample mean X̄ follows a normal distribution with mean μ and standard error σ/√n. We standardise using Z = (X̄ − μ₀) / (σ/√n) to obtain a z-score. The critical region is determined from the standard normal distribution according to α. For a two-tailed test at the 5% level, the critical z-values are ±1.96. If the calculated z falls beyond these, we reject H₀. This is known as a z-test for the mean.

当总体服从正态分布且方差 σ² 已知时，样本均值 X̄ 服从均值为 μ、标准误为 σ/√n 的正态分布。我们使用 Z = (X̄ − μ₀) / (σ/√n) 进行标准化以得到 z 分数。临界区域根据 α 从标准正态分布中确定。对于 5% 水平的双尾检验，临界 z 值为 ±1.96。如果计算出的 z 值超出此范围，我们拒绝 H₀。这称为均值的 z 检验。

---

9. Hypothesis Testing for Correlation Coefficient | 相关系数的假设检验

To test whether a linear relationship exists between two variables, we examine the product moment correlation coefficient r. The null hypothesis H₀: ρ = 0 (no correlation) is tested against H₁: ρ ≠ 0 or a one-tailed alternative. Using the sample size n, we compare r to a critical value from the PMCC table. If |r| exceeds the critical value at 5% significance, we reject H₀ and conclude there is evidence of correlation. This non-parametric test does not require the data to follow a specific distribution.

为了检验两个变量之间是否存在线性关系，我们考察乘积矩相关系数 r。原假设 H₀: ρ = 0（无相关）与备择假设 H₁: ρ ≠ 0 或单尾替代假设进行检验。根据样本量 n，我们将 r 与 PMCC 表中的临界值进行比较。如果 |r| 超过 5% 显著性水平下的临界值，我们拒绝 H₀，得出存在相关的证据。这种非参数检验不要求数据遵循特定分布。

---

10. Using Statistical Tables and Calculators | 使用统计表和计算器

Edexcel examinations require proficiency in reading binomial cumulative distribution tables, normal distribution tables, and PMCC critical value tables. For binomial tests, you will often need to use the ‘less than or equal to’ form to locate probabilities. For normal tests, the percentage points table gives critical z-values for given tail probabilities. Many questions also allow the use of an approved calculator’s distribution functions to find exact p-values. Always sketch a diagram of the distribution to avoid tail errors.

Edexcel 考试要求熟练阅读二项累积分布表、正态分布表和 PMCC 临界值表。对于二项检验，你通常需要使用“小于或等于”形式来查找概率。对于正态检验，百分点表给出了给定尾部概率下的临界 z 值。许多试题也允许使用经批准的计算器分布函数来求得精确的 p 值。始终绘制分布草图以避免尾部错误。

---

11. Step-by-Step Procedure for Hypothesis Testing | 假设检验的步骤分解

A structured approach is vital for securing marks. The typical steps are: 1) State H₀ and H₁ clearly. 2) Specify the significance level α. 3) Identify the test statistic and its distribution under H₀. 4) Determine the critical region or the rule for rejection. 5) Calculate the test statistic from the sample. 6) Compare the statistic to the critical value, or compare the p-value to α. 7) Write a conclusion in context: ‘There is sufficient / insufficient evidence to reject H₀…’ Avoid vague language like ‘accept H₀’; instead say ‘do not reject H₀’.

有条理的方法对于获得分数至关重要。典型步骤如下：1）清楚陈述 H₀ 和 H₁。2）指定显著性水平 α。3）确定检验统计量及其在 H₀ 下的分布。4）确定临界区域或拒绝规则。5）从样本中计算检验统计量。6）将统计量与临界值比较，或将 p 值与 α 比较。7）在上下文中写下结论：“有足够 / 没有足够的证据拒绝 H₀……”避免使用“接受 H₀”这样的模糊语言；而应说“不拒绝 H₀”。

---

12. Common Misconceptions and Exam Tips | 常见误解与应试技巧

Many students confuse the p-value with the significance level: the p-value is computed from the data, while α is chosen beforehand. Another error is to state a conclusion about the sample instead of the population — remember, the test draws conclusions about the population parameter. In binomial tests, always check whether the test is one‑ or two‑tailed before halving significance levels. When a question asks for ‘critical region’, you must give the set of values of the test statistic that reject H₀. Practise past paper questions to become familiar with mark schemes and precise phrasing expected by Edexcel.

许多学生混淆 p 值与显著性水平：p 值由数据计算得出，而 α 是事先选定的。另一个错误是针对样本而非总体陈述结论——记住，检验得出的是关于总体参数的结论。在二项检验中，始终在将显著性水平减半之前检查检验是单尾还是双尾。当题目要求“临界区域”时，你必须给出拒绝 H₀ 的检验统计量取值的集合。通过练习往年试题，熟悉 Edexcel 的评分方案和期望的精确措辞。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）