📚 Significance and Hypothesis Testing | 显著性检验与假设检验
Hypothesis testing lies at the heart of statistical inference, enabling us to use sample data to evaluate claims about population parameters. From determining whether a new drug is effective to checking if a production process is out of control, significance tests help us make decisions in the face of uncertainty. In IB Mathematics, you are expected to understand the logic of hypothesis testing, interpret p-values, recognise errors, and apply tests for means and proportions using appropriate distributions.
假设检验是统计推断的核心，它让我们能够利用样本数据来评估关于总体参数的主张。从判断一种新药是否有效，到检查生产过程是否失控，显著性检验帮助我们在不确定性中做出决策。在 IB 数学课程中，同学们需要理解假设检验的逻辑，解读 p 值，识别错误类型，并使用适当的分布进行均值和比例的检验。
1. Core Concepts of Hypothesis Testing | 假设检验的核心概念
Hypothesis testing begins with a research question translated into two competing statements about a population parameter. We collect a random sample, compute a test statistic, and assess how compatible the data are with a starting assumption. If the sample result is highly unlikely under that assumption, we reject it in favour of the alternative.
假设检验从一个研究问题开始，将其转化为关于总体参数的两个对立陈述。我们收集一个随机样本，计算检验统计量，并评估数据与初始假设的兼容程度。如果在该假设下样本结果极不可能发生，我们就拒绝该假设而接受备择假设。
The decision rule is based on a significance level (α) chosen before the test. Typical values are 0.05, 0.01 or 0.10. The outcome is either 'reject H₀' or 'do not reject H₀' – note that we never 'accept' the null hypothesis in a definitive sense.
决策规则基于事先选定的显著性水平（α），常用值为 0.05、0.01 或 0.10。结果要么是“拒绝 H₀”，要么是“不拒绝 H₀”——注意，我们绝不能以绝对的口吻说“接受”零假设。
2. Null and Alternative Hypotheses | 零假设与备择假设
until evidence suggests otherwise. For a population mean μ, a typical null is H₀: μ = μ₀.
零假设，记作 H₀，是假设没有效应、没有差异或维持现状的陈述。在没有相反证据之前，我们假设它为真。对于总体均值 μ，典型的零假设是 H₀: μ = μ₀。
The alternative hypothesis, H₁ or Hₐ, is what we aim to support. It can be one-sided (μ > μ₀ or μ < μ₀) or two-sided (μ ≠ μ₀). The choice of H₁ determines whether we use a one-tailed or two-tailed test.
备择假设 H₁ 或 Hₐ 是我们试图支持的观点。它可以是单侧的（μ > μ₀ 或 μ < μ₀），也可以是双侧的（μ ≠ μ₀）。H₁ 的选择决定了我们使用单尾检验还是双尾检验。
A consumer group suspects the mean is lower. Then H₀: μ = 1000, H₁: μ < 1000 (one-tailed).
例如，某制造商声称灯泡平均寿命为 1000 小时，消费者组织怀疑实际均值更低。此时 H₀: μ = 1000，H₁: μ < 1000（单尾检验）。
3. Significance Level (α) | 显著性水平 (α)
The significance level α is the probability of rejecting H₀ when it is actually true – a Type I error. It defines how much evidence we require to reject H₀. An α of 0.05 means we are willing to accept a 5% chance of a false positive.
显著性水平 α 是当 H₀ 为真时拒绝它的概率——即第一类错误发生的概率。它定义了我们拒绝 H₀ 所需的证据强度。α = 0.05 表示我们愿意接受 5% 的假阳性风险。
α also sets the rejection region (critical region). If the test statistic falls inside this region, or if the p-value is smaller than α, we reject H₀. Common α values in IB examinations are 0.05, 0.01 and 0.10.
α 还确定了拒绝域（临界域）。如果检验统计量落入该区域，或 p 值小于 α，我们就拒绝 H₀。IB 考试中常用的 α 值为 0.05、0.01 和 0.10。
4. Test Statistics | 检验统计量
A test statistic measures how far the sample estimate is from the null parameter, scaled by its standard error. For a mean with known population standard deviation σ, we use the z-statistic:
检验统计量衡量样本估计值与零假设参数之间的差距，并按其标准误差进行缩放。对于已知总体标准差 σ 的均值，我们使用 z 统计量：
z = (x̄ − μ₀) / (σ/√n)
When σ is unknown, we replace σ with the sample standard deviation s, and the statistic follows a t-distribution with n−1 degrees of freedom:
当 σ 未知时，我们用样本标准差 s 代替 σ，统计量服从自由度为 n−1 的 t 分布：
t = (x̄ − μ₀) / (s/√n)
ions, the test statistic uses the null proportion p₀ and the standard error √(p₀(1−p₀)/n):
对于比例，检验统计量使用零假设的比例 p₀ 和标准误差 √(p₀(1−p₀)/n)：
z = (p̂ − p₀) / √(p₀(1−p₀)/n)
Choosing the correct statistic is essential; IB problems will often specify whether σ is known or provide raw data requiring a t-test.
选择正确的统计量至关重要；IB 题目通常会明确 σ 是否已知，或提供原始数据要求使用 t 检验。
5. Critical Value Approach | 临界值方法
In the critical value approach, we compare the calculated test statistic with a threshold (critical value) that marks the boundary of the rejection region. The critical value depends on α and whether the test is one-tailed or two-tailed.
在临界值方法中，我们将计算出的检验统计量与作为拒绝域边界的临界值进行比较。临界值取决于 α 以及检验是单尾还是双尾。
For a two-tailed z-test at α = 0.05, the critical values are ±1.96. If |z| tailed test with α = 0.05, the critical value is 1.645 (right-tail) or −1.645 (left-tail).
对于 α = 0.05 的双尾 z 检验，临界值为 ±1.96。如果 |z| > 1.96，则拒绝 H₀。对于 α = 0.05 的单尾检验，右侧临界值为 1.645，左侧为 −1.645。
Using t-tables requires the correct degrees of freedom. IB formula booklets provide critical values for common t-distributions, and your calculator can also generate them.
使用 t 分布表时需要正确的自由度。IB 公式手册提供了常用 t 分布的临界值，你的计算器也可以生成这些值。
6. 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 measures the strength of the evidence against H₀: a very small p-value indicates that the observed result is unlikely under the null.
p 值是在 H₀ 为真的前提下，获得与观测到的检验统计量同等极端或更极端结果的概率。它衡量反对 H₀ 的证据强度：极小的 p 值表明在零假设下观测到的结果不太可能发生。
e > α, we do not reject H₀. For a two-tailed test, the p-value accounts for extremes in both directions; many IB calculator functions directly report the p-value.
如果 p 值 ≤ α，则拒绝 H₀；若 p 值 > α，则不拒绝 H₀。对于双尾检验，p 值会考虑两个方向的极端情况；许多 IB 计算器函数可以直接输出 p 值。
ght be 0.033. At α = 0.05, we reject H₀ and conclude the mean differs from the null value.
示例：单样本 t 检验得到 t = 2.34，自由度 15。双尾 p 值可能为 0.033。在 α = 0.05 下，我们拒绝 H₀，并得出均值不同于零假设值的结论。
7. One-tailed vs Two-tailed Tests | 单尾检验与双尾检验
A one-tailed (one-sided) test examines an effect in a specific direction: H₁: μ > μ₀ or μ < μ₀. It allocates all of α to one tail, making it easier to reject H₀ in that direction, but it cannot detect an effect in the opposite direction.
单尾（单侧）检验考察特定方向的效应：H₁: μ > μ₀ 或 μ < μ₀。它将全部 α 分配到一侧尾部，使得在该方向上更容易拒绝 H₀，但无法检测相反方向的效应。
A two-tailed (two-sided) test has H₁: μ ≠ μ₀ and splits α equally between the two tails. This is more conservative for detecting a difference in a predetermined direction, but it can detect any departure from H₀.
双尾（双侧）检验的备择假设为 H₁: μ ≠ μ₀，并将 α 平均分配到两个尾部。对于检测预定方向的差异，这种方法更为保守，但能够检测出任何偏离 H₀ 的情况。
In IB problems, the wording of the question ('greater than', 'different from', 'has increased') guides the choice. Always check whether a directional claim is being made before setting up the hypotheses.
在 IB 题目中，题干的措辞（“大于”、“不同于”、“已增加”）会提示选择。在建立假设之前，务必检查是否传达了方向性的主张。
8. Type I and Type II Errors | 第一类错误和第二类错误
A Type I error occurs when we reject a true null hypothesis. The probability of a Type I error is exactly α, the significance level. A Type II error occurs when we fail to reject a false null hypothesis; its probability is denoted β.
当零假设为真却被拒绝时，发生第一类错误。第一类错误的概率恰好是显著性水平 α。当零假设为假却没有被拒绝时，发生第二类错误，其概率记作 β。
| Decision / 决策 | H₀ True / H₀ 为真 | H₀ False / H₀ 为假 |
| Reject H₀ / 拒绝 H₀ | Type I Error (α) | Correct Decision (Power = 1 − β) |
| Do not reject H₀ / 不拒绝 H₀ | Correct Decision (1 − α) | Type II Error (β) |
Power, defined as 1 − β, is the probability of correctly rejecting a false H₀. It increases with larger sample sizes, larger effect sizes, and higher α. Understanding trade-offs between α and β is a key conceptual point in IB.
检验功效定义为 1 − β，即正确拒绝一个错误的 H₀ 的概率。检验功效随着样本量增加、效应量增大和 α 提高而增强。理解 α 与 β 之间的权衡是 IB 中的一个重要概念点。
9. One-Sample z-Test for a Mean (Known Variance) | 单样本 z 检验（已知方差）
rom a normal distribution (or n is large), we use the z-test. The test statistic is z = (x̄ − μ₀) / (σ/√n). We then compare with critical z-values or find the p-value using the standard normal distribution.
当总体标准差 σ 已知，且样本来自正态分布（或样本量很大）时，我们使用 z 检验。检验统计量为 z = (x̄ − μ₀) / (σ/√n)。随后，我们与临界 z 值进行比较，或利用标准正态分布求 p 值。
ox label states the net weight is 500 g, with σ = 8 g. A random sample of 36 boxes gives x̄ z = (497 − 500)/(8/6) = −2.25. The two-tailed p-value ≈ 0.024. Since p < 0.05, reject H₀.
假设某麦片包装标注重量为 500 g，且 σ = 8 g。随机抽取 36 盒得到 x̄ = 497 g。在 α = 0.05 下检验真实均值是否不同于 500。H₀: μ = 500, H₁: μ ≠ 500。z = (497 − 500)/(8/6) = −2.25。双尾 p 值 ≈ 0.024。因为 p < 0.05，拒绝 H₀。
10. One-Sample t-Test (Unknown Variance) | 单样本 t 检验（未知方差）
the sample standard deviation s. This leads to the t-test, where the test statistic t = (x̄ − μ₀) / (s/√n 1 degrees of freedom. The t-distribution is wider for small ν, reflecting extra uncertainty.
更贴近实际的是，σ 未知，必须由样本标准差 s 来估计。这就引出了 t 检验，其统计量 t = (x̄ − μ₀) / (s/√n) 服从自由度为 ν = n−1 的 t 分布。当 ν 较小时，t 分布更为扁平，反映了额外的不确定性。
he sample is roughly normal or n is sufficiently large (n ≥ 30 often used as a guideline).
随着 n 的增加，t 分布趋近于标准正态分布。IB 提供 t 分布表，或期望你使用图形计算器（GDC）计算 p 值。务必说明自由度，并确认样本近似正态或 n 足够大（通常以 n ≥ 30 为参照）。
Example: a teacher thinks students' average score has improved from a historical mean of 65. A sample of 25 students yields x̄ . t = (68−65)/(10/5) = 1.5, df = 24. p-value ≈ 0.073. Since p > 0.05, we do not reject H₀.
示例：某教师认为学生平均成绩已从历史均值 65 分有所提高。抽取 25 名学生，得到 x̄ = 68，s = 10。在 α = 0.05 下检验：H₀: μ = 65，H₁: μ > 65（单尾）。t = (68−65)/(10/5) = 1.5，自由度 24。p 值 ≈ 0.073。因为 p > 0.05，我们不拒绝 H₀。
11. Relationship Between Confidence Intervals and Hypothesis Tests | 置信区间与假设检验的关系
A two-sided hypothesis test at level α is directly linked to a (1−α)×100% ies inside, we do not reject H₀. This duality provides a visual check of test conclusions.
显著性水平为 α 的双侧假设检验直接与 (1−α)×100% 置信区间相关联。如果零假设参数值落在置信区间之外，我们拒绝 H₀；若在区间内，则不拒绝 H₀。这种对偶关系为检验结论提供了直观的验证方式。
For the earlier cereal example, a 95% confidence interval for μ is 497 ± 1.96×(8/√36) = (494.39, 499.61). Since 500 lies outside the interval, we reject H₀ – matching the p-value result.
对于前面的麦片例子，μ 的 95% 置信区间为 497 ± 1.96×(8/√36) = (494.39, 499.61)。因为 500 位于区间之外，我们拒绝 H₀——这与 p 值的结果一致。
IB questions often ask you to comment on a claim by constructing a confidence interval. This approach is especially useful when the test is two-tailed.
IB 题目经常要求通过构建置信区间来评价某个主张。当检验为双尾时，这种方法尤其有用。
12. Common Pitfalls and Summary | 常见误区与总结
A common misunderstanding is interpreting the p-value as the probability that H₀ is true. The p-value is a conditional probability: P(data | H₀), not P(H₀ | data). Another is treating 'do not reject H₀' as proof that H₀ is true – absence of evidence is not evidence of absence.
一个常见的误解是将 p 值解释为 H₀ 成立的概率。p 值是条件概率：P(数据 | H₀)，而非 P(H₀ | 数据)。另一个误区是将“不拒绝 H₀”当作 H₀ 为真的证据——缺乏证据并不等同于证据不存在。
Always check assumptions: normality, independence, and the correct test statistic. When using a t-test, note that outliers or heavy skewness can invalidate results, especially for small samples. Finally, connect the conclusion back to the context of the problem – a statistical decision must be communicated in plain language.
务必检验假设条件：正态性、独立性以及正确的检验统计量。使用 t 检验时，注意异常值或严重的偏态分布可能使结果失效，尤其是在小样本情况下。最后，要把结论与问题的具体情境相联系——统计决策必须用通俗的语言表述。
Significance and hypothesis testing form a cornerstone of inferential reasoning, blending conceptual understanding with practical computation. Mastering H₀/H₁ setup, p-values, errors, and the choice between z and t procedures will equip you to tackle any IB question with confidence.
显著性检验与假设检验是推断性推理的基石，融合了概念理解与实际计算。掌握 H₀/H₁ 的设定、p 值、错误类型以及 z 与 t 检验的选择，将使你有信心解决任何 IB 题目。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图