Hypothesis Testing and Significance Levels | 假设检验与显著性水平

📚 Hypothesis Testing and Significance Levels | 假设检验与显著性水平

Hypothesis testing is a fundamental statistical method used to make decisions about a population based on sample data. It allows us to assess whether observed results are statistically meaningful or merely due to random chance. In A-Level mathematics, mastering the logic of hypothesis testing, including null and alternative hypotheses, significance levels, and critical regions, is essential for exam success.

假设检验是一种基于样本数据对总体作出判断的基本统计方法。它帮助我们判断观察到的结果是否具有统计意义，还是仅仅由随机波动造成。在 A-Level 数学中，掌握假设检验的逻辑，包括原假设与备择假设、显著性水平以及临界区域，是考试取得高分的关键。

1. Setting Up Hypotheses: Null and Alternative | 建立假设：原假设与备择假设

Every hypothesis test begins with two competing statements about a population parameter. The null hypothesis, denoted H₀, is a statement of “no effect” or “no change” — it represents the status quo or a claimed value we aim to test. The alternative hypothesis, denoted H₁, is the statement we suspect might be true instead, such as a parameter being greater than, less than, or not equal to the claimed value.

每一次假设检验都始于对总体参数的两个相互对立的陈述。原假设（记作 H₀）是“无效”或“无变化”的陈述——它代表现状或我们想要检验的声称值。备择假设（记作 H₁）则是我们怀疑可能成立的另一个陈述，例如参数大于、小于或不等于声称值。

H₀: The parameter equals a specific value, e.g., p = 0.3 or μ = 50.
H₁: The parameter differs from that value in a specified direction, e.g., p > 0.3, p < 0.3, or p ≠ 0.3.

H₀：参数等于某个特定值，例如 p = 0.3 或 μ = 50。
H₁：参数与声称值存在特定方向的差异，例如 p > 0.3、p < 0.3 或 p ≠ 0.3。

In A-Level exam questions, you must first identify which distribution applies — binomial, Poisson, or normal — and clearly state both hypotheses in terms of the population parameter. Marks are often awarded for correct notation, so always write H₀ and H₁ using proper subscript formatting.

在 A-Level 考试题中，首先要判断适用哪种分布——二项分布、泊松分布还是正态分布——并用总体参数清晰地写出两个假设。考试中常因正确使用符号而得分，所以务必用规范的下标格式写出 H₀ 和 H₁。

2. The Significance Level: Measuring Evidence | 显著性水平：衡量证据强度

The significance level, denoted by α (alpha), is the probability of rejecting the null hypothesis when it is actually true. Common significance levels in A-Level mathematics are 5% (α = 0.05), 1% (α = 0.01), and 10% (α = 0.10). This value represents the maximum acceptable probability of making a Type I error — concluding there is an effect when none exists.

显著性水平（记作 α）是指在原假设实际为真时，错误地拒绝它的概率。在 A-Level 数学中，常见的显著性水平是 5%（α = 0.05）、1%（α = 0.01）和 10%（α = 0.10）。该值代表可以接受的犯第一类错误的最大概率——即当实际上不存在效应时，却得出有效应的结论。

P(Reject H₀ | H₀ is true) = α

Think of α as the threshold of “surprisingness”: if the observed test statistic falls within the most extreme α × 100% of the sampling distribution under H₀, we regard the result as statistically significant and reject H₀.

可以把 α 理解为“令人惊讶程度”的门槛：如果观测到的检验统计量落在原假设成立时抽样分布最极端的 α × 100% 区域内，我们就可以认为结果具有统计显著性，从而拒绝 H₀。

显著性水平 α	概率	解读
10%	0.10	较宽松；更易拒绝 H₀
5%	0.05	A-Level 标准默认水平
1%	0.01	非常严格；需要极强证据

When a test is conducted at the 5% significance level, we are accepting a 1-in-20 chance of wrongly rejecting a true null hypothesis. Lowering α makes the test more conservative, requiring stronger evidence before we are willing to reject H₀.

当检验在 5% 显著性水平下进行时，我们接受每 20 次检验中有 1 次错误拒绝真实原假设的风险。降低 α 会使检验更加保守，需要更强的证据才能拒绝 H₀。

3. Test Statistic and the Sampling Distribution | 检验统计量与抽样分布

The test statistic is a value calculated from sample data that we compare against the sampling distribution under H₀. Depending on the context, this may be a binomial count X, a Poisson count, or a standardised normal z-score. The sampling distribution describes all possible values the statistic could take if H₀ were true.

检验统计量是根据样本数据计算出的数值，用于与 H₀ 成立下的抽样分布进行比较。根据具体情境，它可以是二项计数 X、泊松计数或标准化正态 z 分数。抽样分布描述了假设 H₀ 为真时，该统计量可能取到的所有值。

For a binomial test with parameter p, we set H₀: p = p₀ and observe X ~ B(n, p₀) under the null. The test statistic is simply the observed number of successes, X. We then calculate P(X ≥ x) or P(X ≤ x) depending on the direction of H₁.

对于以 p 为参数的二项检验，我们设定 H₀：p = p₀，在原假设下 X ~ B(n, p₀)。检验统计量就是观测到的成功次数 X。随后根据 H₁ 的方向计算 P(X ≥ x) 或 P(X ≤ x)。

X ~ B(n, p₀) under H₀ | Z = (X̄ − μ₀) / (σ / √n) for normal tests

In normal distribution tests where the population variance is known, the standardised test statistic Z follows a standard normal distribution N(0,1). This allows us to compare the z-score directly with critical values such as 1.645 or 1.96.

在总体方差已知的正态分布检验中，标准化检验统计量 Z 服从标准正态分布 N(0,1)。这使我们能够直接将 z 分数与临界值（如 1.645 或 1.96）进行比较。

4. Critical Region and Critical Values | 临界区域与临界值

The critical region (or rejection region) is the set of all values of the test statistic for which we reject H₀. The critical value is the boundary that separates the critical region from the acceptance region. For a given significance level α, the critical region is chosen so that the probability of the test statistic falling in it equals α when H₀ is true.

临界区域（又称拒绝域）是检验统计量所有导致拒绝 H₀ 的取值集合。临界值是划分临界区域与接受区域的边界。对于给定的显著性水平 α，临界区域的选择使得 H₀ 为真时，检验统计量落入其中的概率恰好等于 α。

Suppose X ~ B(20, 0.4) and we test H₁: p > 0.4 at the 5% level. We find the smallest value c such that P(X ≥ c) ≤ 0.05. If P(X ≥ 12) = 0.0565 and P(X ≥ 13) = 0.0210, then the critical region is X ≥ 13 and the critical value is 12.

假设 X ~ B(20, 0.4)，我们在 5% 水平下检验 H₁：p > 0.4。我们找出满足 P(X ≥ c) ≤ 0.05 的最小值 c。如果 P(X ≥ 12) = 0.0565 而 P(X ≥ 13) = 0.0210，那么临界区域就是 X ≥ 13，临界值为 12。

Critical region: X ≥ 13 | P(X ≥ 13) = 0.0210 ≤ 0.05 ✓

Note that the actual probability of rejection (0.0210) is less than the nominal significance level (0.05) because the binomial distribution is discrete. We always choose the critical region such that its probability does not exceed α.

注意，实际拒绝概率（0.0210）小于名义显著性水平（0.05），这是因为二项分布是离散的。我们选择的临界区域概率永远不超过 α。

5. One-Tailed and Two-Tailed Tests | 单侧检验与双侧检验

One-tailed tests are used when the alternative hypothesis specifies a direction: H₁: p > p₀ or H₁: p < p₀. The critical region lies entirely in one tail of the distribution. For a lower-tailed test at the 5% level, we reject H₀ when the observed statistic is among the smallest 5% of values.

单侧检验用于备择假设指明方向的情形：H₁：p > p₀ 或 H₁：p < p₀。临界区域完全位于分布的单侧尾部。对于 5% 水平的下尾检验，当观测统计量落在最小的 5% 取值范围内时，我们拒绝 H₀。

Two-tailed tests are used when H₁ states H₁: p ≠ p₀, without specifying a direction. Here we split α into two equal halves: α/2 in each tail. At the 5% significance level, we put 2.5% in the upper tail and 2.5% in the lower tail.

双侧检验用于 H₁ 为 H₁：p ≠ p₀ 且未指明方向的情形。此时我们将 α 平分为两部分：每侧尾部各占 α/2。在 5% 显著性水平下，上尾和下尾各占 2.5%。

检验类型	备择假设	临界区域（α = 5%）
上尾单侧	p > p₀	上尾 5%
下尾单侧	p < p₀	下尾 5%
双侧	p ≠ p₀	上下尾各 2.5%

For the normal distribution, the critical z-values are ±1.645 for a one-tailed test at 5% and ±1.96 for a two-tailed test at 5%.

对正态分布而言，5% 单侧检验的临界 z 值为 ±1.645，5% 双侧检验的临界 z 值为 ±1.96。

6. The P-Value Approach | P 值方法

The p-value is the probability of obtaining a test statistic at least as extreme as the one observed, assuming H₀ is true. It quantifies how compatible the data are with the null hypothesis. The decision rule is simple: reject H₀ if the p-value is less than or equal to the significance level α; otherwise, do not reject H₀.

P 值是指在原假设 H₀ 为真的前提下，获得与观测值同等或更极端的检验统计量的概率。它量化了数据与原假设的相容程度。决策规则很简单：如果 P 值小于或等于显著性水平 α，则拒绝 H₀；否则不拒绝 H₀。

In a binomial context with observed value x, for H₁: p > p₀, the p-value is P(X ≥ x); for H₁: p < p₀, it is P(X ≤ x); for H₁: p ≠ p₀, it is 2 × min[P(X ≥ x), P(X ≤ x)].

在二项背景下，若观测值为 x：对于 H₁：p > p₀，P 值为 P(X ≥ x)；对于 H₁：p < p₀，P 值为 P(X ≤ x)；对于 H₁：p ≠ p₀，P 值为 2 × min[P(X ≥ x)，P(X ≤ x)]。

Reject H₀ if: p-value ≤ α | Fail to reject H₀ if: p-value > α

The p-value approach and the critical region approach always lead to the same conclusion. However, the p-value provides additional information: a p-value of 0.001 indicates much stronger evidence against H₀ than a p-value of 0.049, even though both lead to rejection at the 5% level.

P 值方法与临界区域方法总会得出相同的结论。然而，P 值提供了更多信息：P 值为 0.001 比 P 值为 0.049 更有力地反驳 H₀，尽管两者在 5% 水平下都导致拒绝。

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

A Type I error occurs when we reject H₀ when it is in fact true. The probability of a Type I error is exactly the significance level α. A Type II error occurs when we fail to reject H₀ when it is actually false. The probability of a Type II error is denoted β (beta).

第一类错误发生在原假设实际为真时我们却拒绝了它。第一类错误的概率恰好等于显著性水平 α。第二类错误发生在原假设实际为假时我们却没有拒绝它。第二类错误的概率记作 β（beta）。

	H₀ 为真	H₀ 为假
拒绝 H₀	第一类错误（概率 = α）	正确决策（功效 = 1 − β）
不拒绝 H₀	正确决策（概率 = 1 − α）	第二类错误（概率 = β）

The power of a test is defined as 1 − β, the probability of correctly rejecting a false null hypothesis. Increasing the sample size n increases the power of a test and decreases the probability of a Type II error.

检验的功效定义为 1 − β，即正确拒绝错误原假设的概率。增大样本量 n 可以提高检验功效并降低第二类错误的概率。

8. Worked Example: Binomial Test | 例题详解：二项检验

Problem (Edexcel style): A coin is believed to be biased towards heads. In 20 tosses, 14 heads are observed. Test at the 5% significance level whether the coin is biased towards heads.

题目（Edexcel 风格）：一枚硬币被认为偏向正面。在 20 次投掷中观察到 14 次正面。在 5% 显著性水平下检验这枚硬币是否偏向正面。

Step 1 — State hypotheses: Let p be the probability of heads. H₀: p = 0.5. H₁: p > 0.5 (one-tailed test, biased towards heads). Under H₀, X ~ B(20, 0.5).

第一步——写出假设：设 p 为出现正面的概率。H₀：p = 0.5。H₁：p > 0.5（单侧检验，偏向正面）。在 H₀ 下，X ~ B(20, 0.5)。

Step 2 — Determine the critical region at α = 0.05: Find c such that P(X ≥ c) ≤ 0.05. Using binomial tables: P(X ≥ 15) = 0.0207 and P(X ≥ 14) = 0.0577. Since 0.0577 > 0.05, the critical region is X ≥ 15.

第二步——确定 α = 0.05 下的临界区域：找到满足 P(X ≥ c) ≤ 0.05 的 c。查二项分布表：P(X ≥ 15) = 0.0207，P(X ≥ 14) = 0.0577。由于 0.0577 > 0.05，临界区域为 X ≥ 15。

Step 3 — Compare: The observed value is X = 14, which does not fall in the critical region X ≥ 15. Alternatively, the p-value is P(X ≥ 14) = 0.0577 > 0.05.

第三步——比较：观测值 X = 14，没有落入临界区域 X ≥ 15。或者，P 值为 P(X ≥ 14) = 0.0577 > 0.05。

Step 4 — Conclusion: There is insufficient evidence at the 5% significance level to reject H₀. We cannot conclude that the coin is biased towards heads.

第四步——结论：在 5% 显著性水平下，没有足够证据拒绝 H₀。我们不能得出硬币偏向正面的结论。

9. Worked Example: Normal Distribution Test | 例题详解：正态分布检验

Problem: A machine fills bags of rice with mean weight 500 g and standard deviation 15 g. A sample of 40 bags is taken and the mean weight is found to be 505 g. Test at the 1% significance level whether the mean weight has increased.

题目：一台机器灌装大米袋，平均重量为 500 g，标准差为 15 g。随机抽取 40 袋测得平均重量为 505 g。在 1% 显著性水平下检验平均重量是否增加了。

Step 1: H₀: μ = 500. H₁: μ > 500. Since n = 40 is large, using the central limit theorem, X̄ ~ N(500, 15²/40).

第一步：H₀：μ = 500。H₁：μ > 500。由于 n = 40 较大，根据中心极限定理，X̄ ~ N(500，15²/40)。

Step 2 — Calculate the test statistic:

Z = (505 − 500) / (15/√40) = 5 / 2.3717 = 2.108

第二步——计算检验统计量：

Z = (505 − 500) / (15/√40) = 5 / 2.3717 = 2.108

Step 3 — Compare with critical value: At α = 0.01 for a one-tailed test, the critical z-value is 2.326. Since 2.108 < 2.326, we do not reject H₀.

第三步——与临界值比较：在 α = 0.01 的单侧检验中，临界 z 值为 2.326。因为 2.108 < 2.326，我们不拒绝 H₀。

Step 4 — Conclusion: At the 1% significance level, there is insufficient evidence to conclude that the mean weight has increased.

第四步——结论：在 1% 显著性水平下，没有足够证据表明平均重量增加了。

10. Common Pitfalls in Hypothesis Testing | 假设检验中的常见误区

One common error is using the wrong alternative hypothesis direction. Read the question carefully: “greater than” indicates an upper-tailed test, “less than” indicates a lower-tailed test, and “different from” or “changed” indicates a two-tailed test.

常见的错误之一是使用的备择假设方向不对。仔细审题：“大于”对应上尾检验，“小于”对应下尾检验，“不同于”或“改变了”对应双侧检验。

Forgetting to state hypotheses in terms of the parameter (p, μ, or λ), not the sample statistic.
Using α wrongly in two-tailed tests: remember to halve α for each tail.
Confusing the critical value with the critical region — the region includes all values beyond the critical value.
Concluding “accept H₀” instead of “fail to reject H₀” — the distinction is subtle but important statistically.
Ignoring the condition np ≥ 5 and n(1−p) ≥ 5 before using a normal approximation to the binomial.

忘记用参数（p、μ 或 λ）来表述假设，而不是用样本统计量。
在双侧检验中错误使用 α：记住每个尾部要用 α/2。
混淆临界值与临界区域——临界区域包含所有超越临界值的取值。
把结论写成“接受 H₀”而不是“不拒绝 H₀”——这一区别细微但在统计上非常重要。
在使用正态近似二项分布前，忽略条件 np ≥ 5 且 n(1−p) ≥ 5。

11. Exam Tips and Conclusion | 考试技巧与总结

In exam questions, always structure your answer systematically: (1) define the parameter and state H₀ and H₁ with correct notation; (2) state the distribution of the test statistic under H₀; (3) determine the critical region or compute the p-value; (4) compare and state your conclusion in the context of the problem.

在考试中，务必系统性地组织答案：（1）定义参数并用规范符号写出 H₀ 和 H₁；（2）写出 H₀ 下检验统计量的分布；（3）确定临界区域或计算 P 值；（4）在题目情境中给出比较和结论。

Remember that a conclusion must always be contextual — write “there is sufficient evidence to suggest that the mean weight has increased” rather than merely “reject H₀”. Examiners reward candidates who link the statistical decision back to the original problem.

记住，结论必须结合题目情境——应该写“有充分证据表明平均重量增加了”，而不是只写“拒绝 H₀”。考官会奖励那些把统计决策与原问题联系起来的考生。

Hypothesis testing is not about proving a claim to be true with absolute certainty; it is about assessing the strength of evidence against a default assumption. The significance level sets the standard of proof, and the p-value quantifies how surprising the data are under H₀.

假设检验并不是要用绝对确定性去证明某个主张为真；而是评估数据对抗默认假设的证据强度。显著性水平设定证明标准，P 值量化数据在原假设下令人惊讶的程度。

This completes our comprehensive revision of hypothesis testing and significance levels. Practise past paper problems on binomial, Poisson, and normal hypotheses to build confidence, and always check whether your test is one-tailed or two-tailed before computing the critical region.

以上就是关于假设检验与显著性水平的全面复习。多练习二项分布、泊松分布和正态分布的真题，逐步建立信心，并且在计算临界区域之前，务必确认检验是单尾还是双尾。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Hypothesis Testing and Significance Levels | 假设检验与显著性水平