Hypothesis Testing with the Normal Distribution | 正态分布中的假设检验

📚 Hypothesis Testing with the Normal Distribution | 正态分布中的假设检验

Hypothesis testing is one of the most powerful tools in A-Level Mathematics, and when combined with the normal distribution, it allows us to make informed decisions about population means using sample data. This article provides a complete, exam-focused guide to conducting hypothesis tests for a normal population mean, including one-tailed and two-tailed tests, critical regions, p-values, and common pitfalls.

假设检验是A-Level数学中最强大的工具之一，当它与正态分布结合时，我们可以利用样本数据对总体均值做出有根据的推断。本文将提供一份完整且紧扣考点的指南，涵盖正态总体均值的假设检验，包括单尾检验、双尾检验、临界区域、p值以及常见易错点。

1. The Normal Distribution and Notation | 正态分布与记号

A continuous random variable \(X\) is said to follow a normal distribution with mean \(\mu\) and variance \(\sigma^2\). We write \(X \sim N(\mu, \sigma^2)\). The standard normal distribution has mean 0 and variance 1, denoted \(Z \sim N(0,1)\). In hypothesis testing, we often standardise a sample mean using the formula:

连续型随机变量 \(X\) 服从均值为 \(\mu\)、方差为 \(\sigma^2\) 的正态分布时，我们记为 \(X \sim N(\mu, \sigma^2)\)。标准正态分布的均值为0、方差为1，记为 \(Z \sim N(0,1)\)。在假设检验中，我们通常使用以下公式将样本均值标准化：

Z = (X̄ − μ₀) / (σ / √n)

where \(X̄\) is the sample mean, \(\mu_0\) is the hypothesised population mean, \(\sigma\) is the population standard deviation, and \(n\) is the sample size. This standardisation converts the sample mean into a z-score, which we can compare against critical values from the standard normal table.

其中 \(X̄\) 是样本均值，\(\mu_0\) 是假设的总体均值，\(\sigma\) 是总体标准差，\(n\) 是样本容量。这一标准化过程将样本均值转换为z分数，我们可以将其与标准正态分布表中的临界值进行比较。

2. The Logic of Hypothesis Testing | 假设检验的逻辑

Hypothesis testing begins with a claim about a population parameter, usually the mean \(\mu\). We set up two competing hypotheses: the null hypothesis \(H_0\) and the alternative hypothesis \(H_1\). The null hypothesis always contains equality, such as \(\mu = \mu_0\), while the alternative hypothesis can be one-tailed (\(\mu < \mu_0\) or \(\mu > \mu_0\)) or two-tailed (\(\mu \neq \mu_0\)).

假设检验始于对总体参数（通常是均值 \(\mu\)）的某种论断。我们建立两个对立的假设：原假设 \(H_0\) 和备择假设 \(H_1\)。原假设总是包含等号，如 \(\mu = \mu_0\)，而备择假设可以是单尾的（\(\mu < \mu_0\) 或 \(\mu > \mu_0\)）或双尾的（\(\mu \neq \mu_0\)）。

We then collect sample data and ask: if \(H_0\) were true, how likely would we be to observe a sample mean at least as extreme as the one we got? If this probability is very small, we conclude that \(H_0\) is unlikely to be true and we reject it in favour of \(H_1\).

然后我们收集样本数据并反问：如果 \(H_0\) 为真，我们观察到与当前样本均值同样极端甚至更极端的样本均值的概率有多大？如果这个概率非常小，我们就认为 \(H_0\) 不太可能成立，从而拒绝 \(H_0\)，转而支持 \(H_1\)。

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

For a normal distribution, the hypotheses are always stated in terms of the population mean \(\mu\). For example, a manufacturer claims that the mean breaking strength of a rope is 500 N. To test whether the rope is actually weaker, we set:

对于正态分布，假设检验总是围绕总体均值 \(\mu\) 来表述。例如，某制造商声称绳索的平均断裂强度为500 N。为了检验绳索是否实际上更弱，我们设定：

H₀: μ = 500 H₁: μ < 500

This is a one-tailed test because the alternative hypothesis specifies a direction. If we only wanted to know whether the mean differs from 500, we would use a two-tailed test:

这是一个单尾检验，因为备择假设指明了方向。如果我们只想了解均值是否不等于500，则应使用双尾检验：

H₀: μ = 500 H₁: μ ≠ 500

In A-Level questions, the wording of the problem usually tells you which alternative hypothesis to use. Phrases like “less than”, “greater than”, “improved”, “reduced” suggest a one-tailed test, while “changed”, “different from” or “not equal to” suggest a two-tailed test.

在A-Level题目中，问题表述通常会提示你应该使用哪种备择假设。诸如“小于”“大于”“提高”“降低”等短语暗示单尾检验，而“改变”“不同于”或“不等于”则暗示双尾检验。

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

The significance level, denoted by \(\alpha\), is the probability of rejecting \(H_0\) when it is actually true. Common significance levels are 5%, 1% and 10%. In a one-tailed test at the 5% level, the critical region occupies the extreme 5% of the standard normal distribution. For \(H_1: \mu > \mu_0\), the critical region is in the upper tail; for \(H_1: \mu < \mu_0\), it is in the lower tail.

显著性水平用 \(\alpha\) 表示，是指当 \(H_0\) 真实成立时却拒绝它的概率。常见的显著性水平为5%、1%和10%。在5%水平的单尾检验中，临界区域占据标准正态分布最极端的5%。对于 \(H_1: \mu > \mu_0\)，临界区域位于上尾；对于 \(H_1: \mu < \mu_0\)，临界区域位于下尾。

For a two-tailed test at the 5% level, the critical region is split equally: 2.5% in the upper tail and 2.5% in the lower tail. The critical values for a 5% two-tailed test are \(z = ±1.96\), while for a 5% one-tailed test the critical value is \(z = 1.645\) (upper) or \(z = -1.645\) (lower).

对于5%水平的双尾检验，临界区域被均分为两部分：上尾2.5%和下尾2.5%。5%双尾检验的临界值为 \(z = ±1.96\)，而5%单尾检验的临界值为 \(z = 1.645\)（上尾）或 \(z = -1.645\)（下尾）。

Common critical values: z₀.₀₅ = 1.645, z₀.₀₂₅ = 1.96, z₀.₀₁ = 2.326, z₀.₀₀₅ = 2.576

If the calculated test statistic falls inside the critical region, we reject \(H_0\). If it falls outside, we do not reject \(H_0\).

如果计算出的检验统计量落在临界区域内，我们就拒绝 \(H_0\)。如果落在临界区域之外，则不能拒绝 \(H_0\)。

5. The Test Statistic | 检验统计量

When the population variance \(\sigma^2\) is known, or the sample size is large enough for the Central Limit Theorem to apply, we use the z-test statistic:

当总体方差 \(\sigma^2\) 已知，或样本容量足够大使得中心极限定理适用时，我们使用z检验统计量：

Z = (X̄ − μ₀) / (σ / √n)

If the population variance is unknown but \(n\) is large (typically \(n \ge 30\)), we may replace \(\sigma\) with the sample standard deviation \(s\). However, in A-Level Edexcel and OCR specifications, hypothesis tests for a normal mean are usually given with a known population variance or a known standard deviation, so this substitution is rarely needed.

如果总体方差未知但样本容量较大（通常 \(n \ge 30\)），我们可以用样本标准差 \(s\) 代替 \(\sigma\)。然而，在Edexcel和OCR的A-Level考纲中，正态均值的假设检验通常直接给出总体方差或标准差，因此很少需要这种替换。

Once the z-score is calculated, we compare it with the critical value. Alternatively, we can compute the p-value, which is the probability of obtaining a test statistic at least as extreme as the observed one under \(H_0\).

计算出z分数后，我们将其与临界值进行比较。或者，我们也可以计算p值，即在 \(H_0\) 成立时得到与观测值同样极端或更极端检验统计量的概率。

6. One-Tailed Test: Worked Example | 单尾检验：例题详解

Example: A machine fills bags of sugar with a mean mass of 1000 g and standard deviation 15 g. A consumer group suspects that the machine is under-filling the bags. A random sample of 25 bags has a mean mass of 993 g. Test at the 5% significance level whether the machine is under-filling.

例题：一台机器灌装糖袋，平均质量为1000 g，标准差为15 g。一个消费者团体怀疑该机器灌装不足。随机抽取25袋，样本平均质量为993 g。在5%显著性水平下检验机器是否灌装不足。

Solution: Let \(X\) be the mass of a bag. We have \(X \sim N(\mu, 15^2)\). The hypotheses are:

解答：设 \(X\) 为一袋糖的质量。已知 \(X \sim N(\mu, 15^2)\)。假设为：

H₀: μ = 1000 H₁: μ < 1000

This is a one-tailed test at the 5% level. The critical value is \(z = -1.645\). The test statistic is:

这是5%水平的单尾检验，临界值为 \(z = -1.645\)。检验统计量为：

Z = (993 − 1000) / (15 / √25) = −7 / 3 = −2.333

Since \(-2.333 < -1.645\), the test statistic lies in the critical region. We reject \(H_0\). There is sufficient evidence at the 5% level to conclude that the machine is under-filling the bags.

由于 \(-2.333 < -1.645\)，检验统计量落在临界区域内。我们拒绝 \(H_0\)。在5%显著性水平下有充分证据表明机器确实灌装不足。

7. Two-Tailed Test: Worked Example | 双尾检验：例题详解

Example: The IQ of students at a school is believed to follow a normal distribution with mean 105 and standard deviation 12. A researcher thinks the mean IQ has changed. A sample of 36 students gives a mean IQ of 109. Test at the 1% significance level whether the mean IQ has changed.

例题：某校学生的智商被认为服从均值为105、标准差为12的正态分布。一位研究者认为平均智商已经发生变化。随机抽取36名学生，样本平均智商为109。在1%显著性水平下检验平均智商是否发生变化。

Solution: The hypotheses are:

解答：假设为：

H₀: μ = 105 H₁: μ ≠ 105

This is a two-tailed test at the 1% level, so each tail has 0.5%. The critical values are \(z = ±2.576\). The test statistic is:

这是1%水平的双尾检验，因此每个尾部分配0.5%。临界值为 \(z = ±2.576\)。检验统计量为：

Z = (109 − 105) / (12 / √36) = 4 / 2 = 2.000

Since \(2.000 < 2.576\), the test statistic does not lie in the critical region. We do not reject \(H_0\). There is insufficient evidence at the 1% level to conclude that the mean IQ has changed.

由于 \(2.000 < 2.576\)，检验统计量没有落在临界区域内。我们不能拒绝 \(H_0\)。在1%显著性水平下没有足够证据表明平均智商发生变化。

8. P-Value Method | p值方法

Instead of comparing a test statistic with a critical value, many exam boards now require or accept the p-value method. The p-value is the probability of obtaining a sample mean at least as extreme as the observed one, assuming \(H_0\) is true. For a one-tailed test, the p-value is the area in one tail; for a two-tailed test, it is the sum of the areas in both tails.

除了将检验统计量与临界值比较之外，许多考试局现在要求或接受p值方法。p值是在假设 \(H_0\) 成立的前提下，得到与观测样本均值同样极端或更极端的样本均值的概率。对于单尾检验，p值是单侧尾部的面积；对于双尾检验，p值是两侧尾部面积之和。

For the one-tailed example above, the test statistic is \(z = -2.333\). The p-value is \(P(Z < -2.333) = 0.0098\). Since \(0.0098 < 0.05\), we reject \(H_0\). For the two-tailed example, the test statistic is \(z = 2.000\). The p-value is \(2 × P(Z > 2.000) = 2 × 0.0228 = 0.0456\). Since \(0.0456 > 0.01\), we do not reject \(H_0\).

对于上面的单尾例题，检验统计量为 \(z = -2.333\)，p值为 \(P(Z < -2.333) = 0.0098\)。因为 \(0.0098 < 0.05\)，我们拒绝 \(H_0\)。对于双尾例题，检验统计量为 \(z = 2.000\)，p值为 \(2 × P(Z > 2.000) = 2 × 0.0228 = 0.0456\)。因为 \(0.0456 > 0.01\)，我们不能拒绝 \(H_0\)。

The decision rule is simple: if p-value < significance level, reject \(H_0\); if p-value ≥ significance level, do not reject \(H_0\).

决策规则很简单：如果p值小于显著性水平，则拒绝 \(H_0\)；如果p值大于或等于显著性水平，则不能拒绝 \(H_0\)。

9. Using Sample Mean or Standardised Variable | 使用样本均值或标准化变量

Some students prefer to conduct the test using the sample mean \(\bar{X}\) directly rather than the z-score. The sampling distribution of \(\bar{X}\) is:

有些学生更喜欢直接用样本均值 \(\bar{X}\) 进行检验，而不是使用z分数。样本均值 \(\bar{X}\) 的抽样分布为：

X̄ ~ N(μ₀, σ²/n) under H₀

We then calculate the critical value for \(\bar{X}\) itself. For a one-tailed test at the 5% level with \(H_1: \mu < \mu_0\), the critical value of \(\bar{X}\) is:

然后我们计算 \(\bar{X}\) 本身的临界值。对于5%水平、 \(H_1: \mu < \mu_0\) 的单尾检验， \(\bar{X}\) 的临界值为：

X̄_c = μ₀ − 1.645 × (σ / √n)

If the observed sample mean is less than this critical value, we reject \(H_0\). Both methods are mathematically equivalent, but the z-score method is usually more compact for exam solutions.

如果观测到的样本均值小于该临界值，我们就拒绝 \(H_0\)。两种方法在数学上是等价的，但z分数方法在考试解答中通常更简洁。

10. Type I and Type II Errors | 第一类错误与第二类错误

A Type I error is the mistake of rejecting \(H_0\) when it is actually true. The probability of a Type I error is exactly the significance level \(\alpha\). A Type II error is the mistake of not rejecting \(H_0\) when it is actually false. The probability of a Type II error is denoted by \(\beta\), and the power of a test is \(1 – \beta\).

第一类错误是指当 \(H_0\) 实际上为真时却拒绝了它。第一类错误的概率恰好等于显著性水平 \(\alpha\)。第二类错误是指当 \(H_0\) 实际上为假时却没有拒绝它。第二类错误的概率记为 \(\beta\)，检验的势为 \(1 – \beta\)。

In A-Level questions, you may be asked to calculate the probability of a Type II error for a specific alternative value of \(\mu\). For example, suppose the true mean is actually 995 g in the sugar bag example. The probability of a Type II error is the probability that the sample mean falls outside the critical region when \(\mu = 995\).

在A-Level题目中，你可能会被要求针对某个具体的备选均值计算第二类错误的概率。例如，假设糖袋例子中的真实均值实际上是995 g。第二类错误的概率就是当 \(\mu = 995\) 时，样本均值落在临界区域之外的概率。

P(Type II error) = P(do not reject H₀ | μ = 995)

To compute this, find the rejection region for \(\bar{X}\) under \(H_0\), then standardise the boundary value using \(\mu = 995\) instead of \(\mu_0\).

计算时，先求出 \(H_0\) 下 \(\bar{X}\) 的拒绝区域，然后用 \(\mu = 995\) 而不是 \(\mu_0\) 来标准化边界值。

11. Common Pitfalls and Exam Tips | 常见易错点与考试技巧

Always state \(H_0\) and \(H_1\) clearly before doing any calculations. Examiners award method marks for correct hypotheses.

在做任何计算之前，务必先清楚地写出 \(H_0\) 和 \(H_1\)。考官会对正确的假设给出方法分。
Check whether the test is one-tailed or two-tailed. Using a one-tailed critical value in a two-tailed test is a very common error.

检查检验是单尾还是双尾。在双尾检验中使用单尾临界值是非常常见的错误。
Remember to divide \(\sigma\) by \(\sqrt{n}\). Forgetting the square root is the most frequent numerical mistake.

记得将 \(\sigma\) 除以 \(\sqrt{n}\)。忘记平方根是最常见的数值错误。
Use the correct notation: \(\bar{X}\) for the sample mean, not \(X\). Write probabilities with correct inequality signs.

使用正确的记号：样本均值为 \(\bar{X}\) 而不是 \(X\)。概率的不等号方向也要写正确。
Conclude in context. Do not just say “reject \(H_0\)” — state what this means about the real-world situation in the question.

结合情境下结论。不要只说“拒绝 \(H_0\)”，还要说明这对题目中的实际情境意味着什么。
If using the p-value method, compare the p-value with the significance level using the correct inequality.

如果使用p值方法，请使用正确的不等号将p值与显著性水平进行比较。

12. Summary | 总结

Hypothesis testing with the normal distribution is a systematic procedure: state the hypotheses, identify the significance level and tail direction, compute the test statistic, compare it with the critical value or p-value, and draw a conclusion in context. Mastery of this topic requires practice with both one-tailed and two-tailed tests, as well as familiarity with critical values and the use of the standard normal table.

正态分布中的假设检验是一个系统化的流程：陈述假设，确定显著性水平和尾部方向，计算检验统计量，将其与临界值或p值进行比较，并结合情境得出结论。掌握这一主题需要大量练习单尾检验和双尾检验，同时熟悉临界值以及标准正态分布表的使用。

By understanding the logic behind the test — not just memorising the steps — you will be able to handle any exam question with confidence, whether it asks for critical regions, p-values, or probabilities of Type II errors.

理解检验背后的逻辑——而不只是死记步骤——你就能自信地应对任何考试题目，无论是要求计算临界区域、p值，还是第二类错误的概率。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Hypothesis Testing with the Normal Distribution | 正态分布中的假设检验