Mastering Statistical Hypothesis Testing for AQA A-Level Mathematics — AQA A-Level 数学统计假设检验完全指南

1. 什么是假设检验？从零假设到结论的完整逻辑链 | What Is Hypothesis Testing? The Complete Logical Chain from Null Hypothesis to Conclusion

假设检验（Hypothesis Testing）是 AQA A-Level 数学统计部分最核心的概念之一，也是 Paper 3 中必考的高分值题型。它的本质是：我们提出一个关于总体参数的声明（claim），然后利用样本数据去判断这个声明是否合理。整个检验过程遵循一套严格的逻辑推理框架，从设立假设（hypotheses）开始，到计算检验统计量（test statistic），再到与临界值（critical value）比较或计算 p 值（p-value），最终得出结论。

Hypothesis testing is one of the most fundamental concepts in the Statistics component of AQA A-Level Mathematics, and it is a guaranteed high-mark question type on Paper 3. At its core, we make a claim about a population parameter and then use sample data to judge whether that claim is reasonable. The entire testing process follows a rigorous logical reasoning framework – from setting up the hypotheses, to calculating the test statistic, to comparing against critical values or computing the p-value, and finally reaching a conclusion.

AQA 考试中要求学生掌握的假设检验类型包括：二项分布（Binomial Distribution）下的假设检验、正态分布（Normal Distribution）下的均值检验，以及相关性与回归的检验。Unit 3 的历年考官报告（examiner report）反复强调，学生在设立备择假设（alternative hypothesis）和正确解释结论时最容易失分。一个常见的错误是草率地写”接受零假设”（accept H₀），而正确的表述应该是”没有足够证据拒绝零假设”（insufficient evidence to reject H₀）。

The types of hypothesis tests required in the AQA specification include: hypothesis testing under the Binomial Distribution, mean testing under the Normal Distribution, and tests for correlation and regression. The Unit 3 examiner reports consistently highlight that students lose the most marks when setting up the alternative hypothesis and when correctly interpreting their conclusion. A common error is writing “accept H₀” – the correct formulation is “there is insufficient evidence to reject H₀.”

2. 零假设与备择假设：单尾检验与双尾检验的判别 | Null and Alternative Hypotheses: Distinguishing One-Tailed and Two-Tailed Tests

假设检验的第一步是设立零假设 H₀ 和备择假设 H₁。零假设是”默认立场” – 通常表示”没有变化””没有差异”或参数等于某个特定值。备择假设则是我们希望找到证据支持的主张。对于总体比例 p 的检验，AQA 标准格式为：H₀: p = p₀ 和 H₁: p < p₀（左尾），p > p₀（右尾），或 p ≠ p₀（双尾）。

The first step in hypothesis testing is to set up the null hypothesis H₀ and the alternative hypothesis H₁. The null hypothesis is the “default position” – typically representing “no change,” “no difference,” or that a parameter equals a specific value. The alternative hypothesis is the claim we hope to find evidence to support. For testing a population proportion p, the AQA standard format is: H₀: p = p₀ and H₁: p < p₀ (left-tailed), p > p₀ (right-tailed), or p ≠ p₀ (two-tailed).

如何判断该用单尾还是双尾？关键看题目措辞。包含”more than””greater than””increased””higher”等词 → 右尾检验（right-tailed）。包含”less than””decreased””lower””fewer”等词 → 左尾检验（left-tailed）。包含”changed””different””not equal”等词 → 双尾检验（two-tailed）。AQA 考官报告指出，许多考生在读到”has the proportion changed?”时错误地使用单尾检验 – 关键词是”changed”，意味着两个方向都有可能，必须用双尾。

How do you decide between one-tailed and two-tailed? The key is in the question wording. Words like “more than,” “greater than,” “increased,” or “higher” indicate a right-tailed test. Words like “less than,” “decreased,” “lower,” or “fewer” indicate a left-tailed test. Words like “changed,” “different,” or “not equal” indicate a two-tailed test. The AQA examiner reports note that many candidates incorrectly use a one-tailed test when reading “has the proportion changed?” – the keyword is “changed,” implying either direction is possible, so a two-tailed test is required.

一个特别需要注意的细节：备择假设必须严格使用不等式。H₁: p > 0.5 是正确的，而 H₁: p ≥ 0.5 是错误的。因为假设检验的逻辑是”尝试推翻零假设”，零假设已经覆盖了等号的情形。如果题目说”test whether the proportion exceeds 0.5″，那么 H₀: p = 0.5，H₁: p > 0.5。

A particularly important detail: the alternative hypothesis must use a strict inequality. H₁: p > 0.5 is correct, while H₁: p ≥ 0.5 is wrong. This is because the logic of hypothesis testing is “attempting to reject the null hypothesis,” and the null hypothesis already covers the equality case. If a question says “test whether the proportion exceeds 0.5,” then H₀: p = 0.5, H₁: p > 0.5.

3. 显著性水平与临界区域：α 值的含义与拒绝域的确定 | Significance Level and Critical Region: The Meaning of α and Determining the Rejection Region

显著性水平 α（significance level）是预先设定的、我们愿意承担的第一类错误（Type I error）的最大概率 – 即当零假设实际为真时，我们错误地拒绝它的概率。AQA A-Level 最常使用的显著性水平是 5%（α = 0.05），偶尔会出现 1% 或 10%。α 的值决定了临界区域的大小：α 越小，拒绝域越窄，要求证据越强才能拒绝 H₀。

The significance level α is the pre-set maximum probability we are willing to accept for a Type I error – that is, the probability of incorrectly rejecting H₀ when it is actually true. AQA A-Level most commonly uses a 5% significance level (α = 0.05), with occasional questions using 1% or 10%. The value of α determines the size of the critical region: a smaller α means a narrower rejection region, requiring stronger evidence to reject H₀.

在二项分布检验中，临界区域（critical region）是一组 X 值（成功次数），当检验统计量落入这个区域时，我们拒绝 H₀。由于二项分布是离散的，实际的显著性水平（actual significance level）往往略低于名义上的 α。AQA 要求学生在答案中明确写出实际的显著性水平。例如：如果 X ~ B(20, 0.5)，对 p > 0.5 做 5% 右尾检验，P(X ≥ 14) = 0.0577 > 0.05，而 P(X ≥ 15) = 0.0207 < 0.05，因此临界区域为 X ≥ 15，实际显著性水平为 2.07%。

In binomial testing, the critical region is a set of X-values (number of successes) such that when the test statistic falls within this region, we reject H₀. Because the binomial distribution is discrete, the actual significance level is often slightly lower than the nominal α. AQA requires candidates to explicitly state the actual significance level in their answer. For example: if X ~ B(20, 0.5), testing p > 0.5 at the 5% level, P(X ≥ 14) = 0.0577 > 0.05, whereas P(X ≥ 15) = 0.0207 < 0.05, so the critical region is X ≥ 15 and the actual significance level is 2.07%.

对于正态分布检验（已知方差 σ²），临界值来自标准正态分布表。在 5% 双尾检验中，临界值为 ±1.96；在 5% 右尾检验中，临界值为 1.6449。AQA 公式册中提供了正态分布百分位表，考生必须熟练使用。常见的 p 值口诀记忆法：”p 值小于 α，拒绝 H₀；p 值大于 α，不拒绝 H₀。”

For normal distribution tests (with known variance σ²), the critical values come from the standard normal distribution table. For a 5% two-tailed test, the critical values are ±1.96; for a 5% right-tailed test, the critical value is 1.6449. The AQA formula booklet provides normal distribution percentage point tables, which candidates must be proficient in using. A common p-value mnemonic: “p-value less than α, reject H₀; p-value greater than α, do not reject H₀.”

4. 二项分布假设检验：用临界区域法逐步求解 | Binomial Hypothesis Testing: Step-by-Step Using the Critical Region Method

二项分布假设检验是 AQA Paper 3 的必考内容，通常出现在试卷前半部分。标准步骤如下：(1) 定义随机变量 X 及其分布，如 X ~ B(n, p)，并明确 n 的含义；(2) 写出 H₀ 和 H₁；(3) 根据 H₁ 的方向确定检验类型（单尾或双尾）；(4) 在 H₀ 为真的假设下，计算临界区域（列出相关的累积概率）；(5) 将观测到的检验统计量与临界区域比较；(6) 写出结论，必须包含上下文（in context），并提及显著性水平。

Binomial hypothesis testing is guaranteed to appear on AQA Paper 3, typically in the first half of the paper. The standard steps are: (1) Define the random variable X and its distribution, e.g. X ~ B(n, p), and clarify what n represents; (2) State H₀ and H₁; (3) Determine the test type (one-tailed or two-tailed) based on the direction of H₁; (4) Assuming H₀ is true, find the critical region by listing relevant cumulative probabilities; (5) Compare the observed test statistic to the critical region; (6) Write a conclusion that must be in context and mention the significance level.

典型例题：一个工厂声称其产品的不合格率为 8%。质检员随机抽取 30 件产品，发现 5 件不合格。在 5% 显著性水平下，检验是否有证据表明不合格率高于声称值。解答：设 X = 不合格产品数量，X ~ B(30, 0.08)。H₀: p = 0.08，H₁: p > 0.08（右尾）。P(X ≥ 5) = 1 − P(X ≤ 4) = 1 − 0.8736 = 0.1264。0.1264 > 0.05，因此 X = 5 不在临界区域内。结论：没有足够证据在 5% 显著性水平下拒绝 H₀，即没有足够证据表明不合格率高于 8%。

Typical exam question: A factory claims its defect rate is 8%. A quality inspector randomly selects 30 items and finds 5 are defective. Test, at the 5% significance level, whether there is evidence that the defect rate exceeds the claimed value. Solution: Let X = number of defective items, X ~ B(30, 0.08). H₀: p = 0.08, H₁: p > 0.08 (right-tailed). P(X ≥ 5) = 1 − P(X ≤ 4) = 1 − 0.8736 = 0.1264. 0.1264 > 0.05, so X = 5 is not in the critical region. Conclusion: There is insufficient evidence, at the 5% significance level, to reject H₀ – that is, insufficient evidence that the defect rate exceeds 8%.

对于双尾二项分布检验，由于分布的离散性，需要将 α 对半分配到两个尾部。但实际操作中通常直接求 p 值的两倍与 α 比较，或者分别检查两个尾部。若 H₁: p ≠ 0.5，X ~ B(20, 0.5)，观测值为 x = 15，则 p 值 = 2 × P(X ≥ 15) = 2 × 0.0207 = 0.0414。因为 0.0414 < 0.05，拒绝 H₀。

For two-tailed binomial tests, because of the distribution’s discreteness, α is split equally between the two tails. In practice, you typically double the p-value and compare against α, or check both tails separately. If H₁: p ≠ 0.5, X ~ B(20, 0.5), and the observed value is x = 15, then p-value = 2 × P(X ≥ 15) = 2 × 0.0207 = 0.0414. Since 0.0414 < 0.05, we reject H₀.

5. 正态分布均值检验：已知方差下用 CLT 和 Z 检验 | Normal Distribution Mean Testing: Z-Tests with Known Variance Using the Central Limit Theorem

当样本量足够大（通常 n ≥ 30）或总体本身服从正态分布时，样本均值的抽样分布近似为正态分布。这是中心极限定理（Central Limit Theorem, CLT）的直接应用。AQA 考试中，正态分布假设检验通常给出总体方差 σ² 或要求从数据中计算。检验统计量公式为：Z = (x̄ − μ₀) / (σ / √n)，其中 x̄ 为样本均值，μ₀ 为零假设下的总体均值，σ 为总体标准差，n 为样本量。

When the sample size is sufficiently large (typically n ≥ 30) or the population itself follows a normal distribution, the sampling distribution of the sample mean is approximately normal. This is a direct application of the Central Limit Theorem (CLT). In AQA exams, normal distribution hypothesis tests typically give the population variance σ² or require it to be calculated from the data. The test statistic formula is: Z = (x̄ − μ₀) / (σ / √n), where x̄ is the sample mean, μ₀ is the population mean under H₀, σ is the population standard deviation, and n is the sample size.

例题：一家食品厂声称其袋装薯片的平均重量为 200 g，标准差为 4 g。质检员随机抽取 25 袋，测得平均重量为 198.3 g。在 1% 显著性水平下，检验平均重量是否发生了变化。解答：H₀: μ = 200，H₁: μ ≠ 200（双尾，因为”发生变化”）。Z = (198.3 − 200) / (4 / √25) = −1.7 / 0.8 = −2.125。双尾 1% 临界值为 ±2.576。|Z| = 2.125 < 2.576，因此不拒绝 H₀。结论：在 1% 显著性水平下，没有足够证据表明袋装薯片的平均重量与声称的 200 g 有差异。

Example: A food factory claims its bags of crisps weigh 200 g on average, with a standard deviation of 4 g. A quality inspector randomly selects 25 bags and finds a mean weight of 198.3 g. Test, at the 1% significance level, whether the mean weight has changed. Solution: H₀: μ = 200, H₁: μ ≠ 200 (two-tailed, because “changed”). Z = (198.3 − 200) / (4 / √25) = −1.7 / 0.8 = −2.125. The 1% two-tailed critical value is ±2.576. |Z| = 2.125 < 2.576, so do not reject H₀. Conclusion: At the 1% significance level, there is insufficient evidence that the mean weight of the crisp bags differs from the claimed 200 g.

需要注意的是，AQA 有时会考察未知方差的情况。此时使用 t 分布（Student’s t-distribution），检验统计量变为 T = (x̄ − μ₀) / (s / √n)，其中 s 为样本标准差。自由度 ν = n − 1。t 检验的临界值比 Z 检验更宽（t 分布的尾部更厚），反映了估计 σ 带来的额外不确定性。

It is worth noting that AQA occasionally examines the case where the variance is unknown. In that case, the t-distribution (Student’s t-distribution) is used, and the test statistic becomes T = (x̄ − μ₀) / (s / √n), where s is the sample standard deviation. The degrees of freedom are ν = n − 1. The critical values for the t-test are wider than those for the Z-test (the t-distribution has thicker tails), reflecting the additional uncertainty from estimating σ.

6. 第一类错误与第二类错误：考官报告中最常失分的概念 | Type I and Type II Errors: The Most Commonly Misunderstood Concept in Examiner Reports

第一类错误（Type I error）和第二类错误（Type II error）是 AQA Unit 3 考官报告中反复提及的高频失分点。第一类错误：当 H₀ 实际为真时，我们却拒绝了 H₀。其概率等于显著性水平 α。第二类错误：当 H₀ 实际为假时，我们却没有拒绝 H₀。其概率记为 β，它依赖于真实的参数值。

Type I error and Type II error are high-frequency points of mark loss repeatedly mentioned in the AQA Unit 3 examiner reports. Type I error: rejecting H₀ when H₀ is actually true. Its probability equals the significance level α. Type II error: failing to reject H₀ when H₀ is actually false. Its probability is denoted β and depends on the true parameter value.

四个关键概念的关系可以总结为：检验的功效（power）= 1 − β = 当 H₀ 为假时正确拒绝它的概率。AQA 经常以文字题形式考察：”解释在这个假设检验中，第一类错误的含义。”标准回答模式为：”第一类错误意味着，实际上不合格率确实是 8%（H₀ 为真），但基于样本数据，我们错误地得出结论认为不合格率高于 8%。这种错误发生的概率是 5%（即显著性水平）。”

The relationship among the four key concepts can be summarised as: power of the test = 1 − β = the probability of correctly rejecting H₀ when it is false. AQA frequently examines this in worded form: “Explain, in the context of this hypothesis test, what a Type I error would mean.” The standard answer pattern is: “A Type I error would mean that, in reality, the defect rate is indeed 8% (H₀ is true), but based on the sample data, we incorrectly conclude that the defect rate exceeds 8%. The probability of this error occurring is 5% (the significance level).”

常见误区：很多学生混淆了 α 和 β。α 由研究者预先设定（通常是 0.05），而 β 取决于样本量 n、真实的参数值以及 α 的选择。增大样本量可以同时降低 α 和 β，这是在实验设计中非常实用的策略。AQA 考试中典型的增大功效的方法题答案为：增加样本量（increase sample size）或增大显著性水平（increase significance level），其中前者是更优的答案，因为后者会增加第一类错误的风险。

A common misconception: many students confuse α and β. α is pre-set by the researcher (usually 0.05), whereas β depends on the sample size n, the true parameter value, and the choice of α. Increasing the sample size can reduce both α and β simultaneously – a very practical strategy in experimental design. The typical AQA exam answer for how to increase power is: increase the sample size or increase the significance level – the former is the superior answer because the latter increases the risk of a Type I error.

7. 相关性与回归的假设检验：PMCC 显著性检验 | Hypothesis Testing for Correlation and Regression: Testing the Significance of PMCC

AQA A-Level 还要求掌握积矩相关系数（Product Moment Correlation Coefficient, PMCC）的显著性检验。零假设为 H₀: ρ = 0（总体中不存在线性相关），备择假设为 H₁: ρ > 0（正相关），ρ < 0（负相关）或 ρ ≠ 0（存在相关）。检验统计量为样本的 PMCC（r），临界值查 PMCC 临界值表（AQA 公式册中提供）。

AQA A-Level also requires mastery of significance testing for the Product Moment Correlation Coefficient (PMCC). The null hypothesis is H₀: ρ = 0 (no linear correlation in the population), and the alternative hypothesis is H₁: ρ > 0 (positive correlation), ρ < 0 (negative correlation), or ρ ≠ 0 (correlation exists). The test statistic is the sample PMCC (r), and the critical value is found from the PMCC critical value table (provided in the AQA formula booklet).

PMCC 临界值表的使用方法：根据样本量 n 和显著性水平 α（单尾或双尾）查找对应的临界值。若 |r| 大于临界值，则拒绝 H₀，认为存在显著的线性相关。例如：n = 12，双尾 5% 检验，查表得临界值为 0.5760。如果计算出的 r = 0.683，因为 0.683 > 0.5760，拒绝 H₀，结论为”在 5% 显著性水平下，有证据表明两个变量之间存在线性相关”。

Using the PMCC critical value table: look up the critical value corresponding to the sample size n and significance level α (one-tailed or two-tailed). If |r| exceeds the critical value, reject H₀ and conclude that a significant linear correlation exists. For example: n = 12, two-tailed 5% test, the table gives a critical value of 0.5760. If the calculated r = 0.683, since 0.683 > 0.5760, reject H₀ and conclude: “At the 5% significance level, there is evidence of a linear correlation between the two variables.”

一个重要提醒：显著性检验通过只表明存在线性相关，并不代表因果关系（causation）。AQA 经常在后续小问中考察这个区别。即使 r 非常接近 1 或 −1，也不能直接声称一个变量的变化导致了另一个变量的变化 – 只能说明它们之间存在强烈的线性关联。混淆相关与因果是 GCSE 到 A-Level 过渡中最需要纠正的统计思维误区之一。

An important reminder: passing a significance test only indicates the existence of a linear correlation, not causation. AQA often examines this distinction in follow-up questions. Even if r is very close to 1 or −1, you cannot directly claim that changes in one variable cause changes in the other – you can only state that there is a strong linear association between them. Confusing correlation with causation is one of the most important statistical thinking misconceptions to correct in the transition from GCSE to A-Level.

8. 从考官报告中学习：答题规范与常见扣分陷阱 | Learning from the Examiner Report: Answer Conventions and Common Deduction Traps

综合 AQA A-Level 数学 Unit 3 多年考官报告，以下是最常见的答题失误，每一条都值得在考前反复检查：

Synthesising multiple years of AQA A-Level Mathematics Unit 3 examiner reports, here are the most common answer mistakes – every one of them is worth double-checking before the exam:

第一，忘记在结论中引用显著性水平。只写”拒绝 H₀”是不够的 – 必须写”在 5% 显著性水平下，拒绝 H₀”。这个短语值一分。第二，p 值比较时忘记说明比较对象。正确的格式是：”0.0207 < 0.05，因此拒绝 H₀”，而不是仅仅写”0.0207，拒绝 H₀”。第三，在双尾检验中使用单尾临界值。第四，二项分布中错误地使用正态近似（除非题目明确要求或 n 足够大使得 np > 5 且 n(1 − p) > 5）。第五，写”接受 H₀”而不是”不拒绝 H₀”。

First, forgetting to reference the significance level in the conclusion. Writing “reject H₀” is insufficient – you must write “reject H₀ at the 5% significance level.” This phrase is worth one mark. Second, failing to state what is being compared when using a p-value. The correct format is: “0.0207 < 0.05, therefore reject H₀,” not simply “0.0207, reject H₀.” Third, using one-tailed critical values in a two-tailed test. Fourth, incorrectly using the normal approximation in a binomial test (unless explicitly required by the question or n is large enough that np > 5 and n(1 − p) > 5). Fifth, writing “accept H₀” instead of “do not reject H₀.”

第六，在上下文结论中缺少”evidence”一词。AQA 评分方案明确要求结论中使用”evidence”或”sufficient evidence”这一关键词。标准句型：”There is sufficient/insufficient evidence, at the X% significance level, to suggest that …”。第七，计算临界区域时忘记注明实际显著性水平。第八，混淆样本统计量（sample statistic，如 x̄）与总体参数（population parameter，如 μ）。

Sixth, omitting the word “evidence” in the contextual conclusion. The AQA mark scheme explicitly requires the keyword “evidence” or “sufficient evidence” in the conclusion. The standard sentence pattern is: “There is sufficient/insufficient evidence, at the X% significance level, to suggest that …” Seventh, forgetting to state the actual significance level when calculating the critical region. Eighth, confusing the sample statistic (e.g. x̄) with the population parameter (e.g. μ).

9. 综合应用题：假设检验在真实情境中的完整推理链 | Integrated Application: The Complete Reasoning Chain of Hypothesis Testing in Real-World Contexts

让我们通过一道 AQA 风格的综合性题目来整合以上所有知识点。题目：一家制药公司声称其新药的治愈率为 70%。在针对 50 名患者的临床试验中，有 42 人被治愈。(a) 在 5% 显著性水平下，检验是否有证据表明治愈率高于声称值。(b) 解释在这个情境中第一类错误的含义。(c) 如果真实的治愈率实际上是 80%，计算第二类错误的概率。

Let us integrate all the above knowledge through an AQA-style comprehensive problem. Question: A pharmaceutical company claims its new drug has a cure rate of 70%. In a clinical trial of 50 patients, 42 are cured. (a) Test, at the 5% significance level, whether there is evidence that the cure rate exceeds the claimed value. (b) Explain what a Type I error would mean in this context. (c) If the true cure rate is actually 80%, calculate the probability of a Type II error.

(a) 解答：X = 治愈人数，X ~ B(50, 0.7)。H₀: p = 0.7，H₁: p > 0.7（右尾检验）。我们需要找到 P(X ≥ 42)。使用二项分布公式或统计表：P(X ≥ 42) = 1 − P(X ≤ 41)。通过二项累积分布，P(X ≤ 41) ≈ 0.9738，因此 P(X ≥ 42) = 1 − 0.9738 = 0.0262。0.0262 < 0.05，因此拒绝 H₀。结论：在 5% 显著性水平下，有足够证据表明该新药的治愈率高于 70%。

(a) Solution: X = number of cured patients, X ~ B(50, 0.7). H₀: p = 0.7, H₁: p > 0.7 (right-tailed test). We need to find P(X ≥ 42). Using the binomial distribution formula or statistical tables: P(X ≥ 42) = 1 − P(X ≤ 41). From the binomial cumulative distribution, P(X ≤ 41) ≈ 0.9738, so P(X ≥ 42) = 1 − 0.9738 = 0.0262. 0.0262 < 0.05, therefore reject H₀. Conclusion: At the 5% significance level, there is sufficient evidence to suggest that the new drug’s cure rate exceeds 70%.

(b) 解答：第一类错误意味着，实际上该药物的真实治愈率确实是 70%（H₀ 为真），但我们基于这 50 名患者的样本数据错误地得出了治愈率高于 70% 的结论。发生这种错误的概率是 5%（即检验的显著性水平）。

(b) Solution: A Type I error would mean that, in reality, the drug’s true cure rate is indeed 70% (H₀ is true), but based on the sample data from these 50 patients, we incorrectly conclude that the cure rate exceeds 70%. The probability of this error occurring is 5% (the significance level of the test).

(c) 解答：当真实治愈率 p = 0.8 时，X ~ B(50, 0.8)。第二类错误发生在 H₀ 为假但我们没有拒绝 H₀ 的情况下 – 即检验统计量没有落入临界区域。首先确定临界区域：在 H₀: p = 0.7 下，5% 右尾临界区域满足 P(X ≥ c) ≤ 0.05。经过计算，c = 42（因为 P(X ≥ 42) = 0.0262 ≤ 0.05 而 P(X ≥ 41) ≈ 0.0529 > 0.05）。当 p = 0.8 时，β = P(不拒绝 H₀ | p = 0.8) = P(X ≤ 41 | p = 0.8)。使用 B(50, 0.8)：P(X ≤ 41) ≈ 0.7123。因此，当真实治愈率为 80% 时，第二类错误概率约为 0.7123。检验功效 = 1 − 0.7123 = 0.2877。

(c) Solution: When the true cure rate is p = 0.8, X ~ B(50, 0.8). A Type II error occurs when H₀ is false but we fail to reject H₀ – that is, the test statistic does not fall in the critical region. First, determine the critical region: under H₀: p = 0.7, the 5% right-tailed critical region satisfies P(X ≥ c) ≤ 0.05. By calculation, c = 42 (since P(X ≥ 42) = 0.0262 ≤ 0.05 while P(X ≥ 41) ≈ 0.0529 > 0.05). When p = 0.8, β = P(do not reject H₀ | p = 0.8) = P(X ≤ 41 | p = 0.8). Using B(50, 0.8): P(X ≤ 41) ≈ 0.7123. Therefore, when the true cure rate is 80%, the probability of a Type II error is approximately 0.7123. The power of the test = 1 − 0.7123 = 0.2877.

10. 考试策略：AQA Paper 3 假设检验题的时间分配与检查清单 | Exam Strategy: Time Management and Checklist for AQA Paper 3 Hypothesis Testing Questions

AQA Paper 3 通常包含 2-3 道假设检验相关题目，分布在试卷的不同位置。前部的二项分布检验通常花费 6-8 分钟，中部的大题（有时涉及正态分布或 PMCC）花费 10-12 分钟。建议的做题检查清单：✓ 随机变量定义了吗？✓ 假设写对了吗（H₀ 总是带等号）？✓ 单尾还是双尾？✓ 临界值/p 值与 α 的比较写清楚了吗？✓ 结论是否包含”in context”+”significance level”+”evidence”三个要素？✓ 实际显著性水平标注了吗？

AQA Paper 3 typically contains 2-3 hypothesis-testing-related questions, distributed across different parts of the paper. A binomial test near the beginning typically takes 6-8 minutes, while a larger question in the middle (sometimes involving normal distribution or PMCC) takes 10-12 minutes. A recommended answer checklist: ✓ Is the random variable defined? ✓ Are the hypotheses correctly stated (H₀ always includes equality)? ✓ One-tailed or two-tailed? ✓ Is the comparison between critical value/p-value and α clearly stated? ✓ Does the conclusion contain all three elements: “in context” + “significance level” + “evidence”? ✓ Is the actual significance level noted?

最明智的复习策略：将近五年的 AQA 真题中的假设检验题全部独立做完，然后对照评分方案逐字逐句核对结论用语。你会发现，结论部分的措辞有非常固定的模板，熟练掌握后可以保证这部分不丢分。同时，不要忽视 AQA 真题中偶尔出现的”根据假设检验结果提出建议”类的延伸问题 – 这类题目考察的是统计推理在实际决策中的应用，需要用非技术语言清晰表达。

The wisest revision strategy: independently complete every hypothesis testing question from the last five years of AQA past papers, then compare your answer against the mark scheme word for word, focusing on the conclusion wording. You will find that the phrasing of the conclusion follows a very fixed template – mastering this ensures you never lose marks on this section. Also, do not overlook the occasional extension questions in AQA papers that ask “based on the hypothesis test result, make a recommendation” – these examine the application of statistical reasoning to real-world decision-making and require clear expression in non-technical language.

---

Summary | 总结

假设检验是 AQA A-Level 数学统计部分的核心技能，贯穿二项分布、正态分布和相关分析三大模块。掌握它的关键在于理解假设检验的逻辑框架而非死记硬背公式：从设立 H₀ 和 H₁ 开始，到判断检验方向，再到计算临界区域或 p 值，最后用规范的语言写出上下文结论。AQA 考官报告反复提醒的要点包括：区分”接受 H₀”与”不拒绝 H₀”、结论中必须包含显著性水平和”evidence”一词、正确识别单尾与双尾检验、以及准确解释第一类错误和第二类错误的实际含义。通过系统的真题训练和规范化的答题模板，假设检验可以成为 AQA Paper 3 中最有把握的得分板块。

Hypothesis testing is a core skill in the Statistics component of AQA A-Level Mathematics, spanning the three major areas of binomial distribution, normal distribution, and correlation analysis. The key to mastery lies in understanding the logical framework of hypothesis testing rather than memorising formulae: from setting up H₀ and H₁, to determining the direction of the test, to calculating the critical region or p-value, and finally writing a contextual conclusion in standardised language. The key points repeatedly highlighted by AQA examiner reports include: distinguishing “accept H₀” from “do not reject H₀,” ensuring the significance level and the word “evidence” appear in the conclusion, correctly identifying one-tailed versus two-tailed tests, and accurately explaining the practical meaning of Type I and Type II errors. Through systematic past-paper practice and standardised answer templates, hypothesis testing can become the most secure mark-scoring section in AQA Paper 3.

更多咨询请联系16621398022（同微信）