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

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

Hypothesis testing is one of the most powerful tools in statistics, allowing us to make data-driven decisions under uncertainty. In this article, we will break down the logic of hypothesis testing, explain the meaning of significance levels, and walk through worked examples suitable for IB Mathematics students.

假设检验是统计学中最强大的工具之一，它让我们在不确定性的情况下做出基于数据的决策。本文将深入剖析假设检验的逻辑，解释显著性水平的含义，并为 IB 数学学生提供完整的例题解析。

1. What Is a Hypothesis Test? | 什么是假设检验？

A hypothesis test is a formal procedure used to decide whether a claim about a population parameter is supported by sample data. We start with two competing statements: the null hypothesis and the alternative hypothesis.

假设检验是一种正式的程序，用于判断关于总体参数的某个主张是否得到样本数据的支持。我们从两个相互对立的陈述开始：原假设和备择假设。

The null hypothesis (denoted H₀) is a statement of no effect, no difference, or no change. It is assumed to be true until evidence suggests otherwise. The alternative hypothesis (denoted H₁ or Hₐ) is the claim we are trying to find evidence for.

原假设（记为 H₀）是表示无效应、无差异或无变化的陈述。在没有相反证据之前，我们假定它为真。备择假设（记为 H₁ 或 Hₐ）则是我们试图寻找证据支持的主张。

For example, if a drug company claims that a new medicine reduces blood pressure, the null hypothesis might be “the medicine has no effect on blood pressure,” while the alternative hypothesis would be “the medicine reduces blood pressure.”

例如，如果一家制药公司声称新药能降低血压，那么原假设可能是”该药对血压没有影响”，而备择假设则是”该药能降低血压”。

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

Hypotheses are always stated about population parameters, not sample statistics. Common parameters include the population mean μ, the population proportion p, and the population variance σ².

假设总是关于总体参数的陈述，而不是关于样本统计量。常见的参数包括总体均值 μ、总体比例 p 和总体方差 σ²。

The null hypothesis always contains an equality sign. Depending on the research question, it can be written as H₀: μ = μ₀, H₀: μ ≤ μ₀, or H₀: μ ≥ μ₀. The alternative hypothesis can be one-tailed (μ < μ₀ or μ > μ₀) or two-tailed (μ ≠ μ₀).

原假设总是包含等号。根据研究问题，它可以写成 H₀: μ = μ₀、H₀: μ ≤ μ₀ 或 H₀: μ ≥ μ₀。备择假设可以是单尾的（μ < μ₀ 或 μ > μ₀），也可以是双尾的（μ ≠ μ₀）。

Test Type	Null Hypothesis	Alternative Hypothesis	When to Use
Two-tailed	H₀: μ = μ₀	H₁: μ ≠ μ₀	We only want to detect a difference, not its direction.
Left-tailed	H₀: μ ≥ μ₀	H₁: μ < μ₀	We suspect the parameter has decreased.
Right-tailed	H₀: μ ≤ μ₀	H₁: μ > μ₀	We suspect the parameter has increased.

H₀: μ = μ₀ vs H₁: μ ≠ μ₀ (two-tailed)

H₀: μ ≥ μ₀ vs H₁: μ < μ₀ (left-tailed)

H₀: μ ≤ μ₀ vs H₁: μ > μ₀ (right-tailed)

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

Once the hypotheses are set, we collect a sample and compute a test statistic. The test statistic measures how far the sample result is from what the null hypothesis predicts, measured in standard errors.

确定假设后，我们收集样本并计算检验统计量。检验统计量衡量样本结果与原假设预测值之间的差距，以标准误差为单位。

For a population mean with known variance σ², the test statistic is:

对于总体方差 σ² 已知的总体均值，检验统计量为：

z = (x̄ − μ₀) / (σ / √n)

where x̄ is the sample mean, μ₀ is the hypothesized population mean, σ is the population standard deviation, and n is the sample size. This z-statistic follows a standard normal distribution under H₀.

其中 x̄ 是样本均值，μ₀ 是假设的总体均值，σ 是总体标准差，n 是样本容量。在原假设成立时，该 z 统计量服从标准正态分布。

When σ is unknown and the sample size is small, we use the t-statistic:

当 σ 未知且样本量较小时，我们使用 t 统计量：

t = (x̄ − μ₀) / (s / √n)

where s is the sample standard deviation. The t-statistic follows a t-distribution with n − 1 degrees of freedom.

其中 s 是样本标准差。t 统计量服从自由度为 n − 1 的 t 分布。

4. Significance Level α | 显著性水平 α

The significance level, denoted by α (alpha), is the probability of rejecting the null hypothesis when it is actually true. This is known as a Type I error. Common choices are α = 0.05, α = 0.01, and α = 0.10.

显著性水平，记为 α（alpha），是在原假设实际为真时拒绝原假设的概率。这被称为第一类错误。常用的选择是 α = 0.05、α = 0.01 和 α = 0.10。

In IB Mathematics, α is usually given in the question. It defines the boundary of the rejection region. If the p-value is less than α, we reject H₀. If the p-value is greater than or equal to α, we fail to reject H₀.

在 IB 数学中，α 通常在题目中给出。它定义了拒绝域的边界。如果 p 值小于 α，我们拒绝 H₀。如果 p 值大于或等于 α，则不能拒绝 H₀。

Think of α as the tolerance for making a false alarm. A smaller α means we require stronger evidence before declaring a result statistically significant.

可以把 α 理解为对”误报”的容忍度。α 越小，意味着我们在宣布结果具有统计显著性之前需要更强的证据。

5. p-Value and Critical Values | p 值与临界值

There are two equivalent ways to make a decision in hypothesis testing: the p-value method and the critical value method.

在假设检验中有两种等价的决策方法：p 值法和临界值法。

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 evidence against H₀.

p 值是在假设 H₀ 为真的前提下，获得至少与观测到的检验统计量一样极端的结果的概率。它衡量反对 H₀ 的证据强度。

For a two-tailed z-test, if z = 2.00, the p-value is calculated as:

对于双尾 z 检验，如果 z = 2.00，p 值的计算方式为：

p-value = 2 × P(Z > 2.00) = 2 × (1 − Φ(2.00)) = 2 × 0.0228 = 0.0456

Since 0.0456 < 0.05, we would reject H₀ at the 5% significance level.

由于 0.0456 < 0.05，我们会在 5% 的显著性水平下拒绝 H₀。

The critical value approach uses the sampling distribution of the test statistic. For a two-tailed test at α = 0.05, the critical z-values are ±1.96. If the observed z falls beyond these values, we reject H₀.

临界值法使用检验统计量的抽样分布。对于 α = 0.05 的双尾检验，临界 z 值为 ±1.96。如果观测到的 z 落在这个范围之外，我们就拒绝 H₀。

6. One-Tailed vs Two-Tailed Tests | 单尾检验与双尾检验

The choice between a one-tailed and two-tailed test depends entirely on the research question. A two-tailed test checks for any difference, while a one-tailed test checks for a difference in a specific direction.

选择单尾检验还是双尾检验完全取决于研究问题。双尾检验检测任何方向的差异，而单尾检验只检测特定方向的差异。

For a two-tailed test at α = 0.05, the total rejection region of 0.05 is split equally into two tails of 0.025 each. For a one-tailed test, the entire 0.05 is placed in one tail.

对于 α = 0.05 的双尾检验，总拒绝域 0.05 被平均分成两个尾部，每个尾部 0.025。对于单尾检验，整个 0.05 都放在一个尾部。

Test Type	Critical z at α = 0.05	Rejection Region
Two-tailed	±1.96	z < −1.96 or z > 1.96
Left-tailed	−1.645	z < −1.645
Right-tailed	1.645	z > 1.645

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

No hypothesis test is perfect. There are two types of errors we can make, and understanding them is essential for interpreting results correctly.

任何假设检验都不是完美的。我们可能犯两类错误，理解它们对于正确解读结果至关重要。

Type I error: Rejecting H₀ when H₀ is actually true. The probability of a Type I error is exactly α, the significance level.

第一类错误：在原假设 H₀ 实际为真时拒绝了 H₀。犯第一类错误的概率正好是 α，即显著性水平。

Type II error: Failing to reject H₀ when H₀ is actually false. The probability of a Type II error is denoted by β. The power of a test is 1 − β, which is the probability of correctly rejecting a false H₀.

第二类错误：在原假设 H₀ 实际为假时未能拒绝 H₀。犯第二类错误的概率记为 β。检验的功效是 1 − β，即正确拒绝虚假 H₀ 的概率。

	H₀ is true	H₀ is false
Reject H₀	Type I error (α)	Correct decision (1 − β)
Fail to reject H₀	Correct decision (1 − α)	Type II error (β)

8. Steps for Conducting a Hypothesis Test | 进行假设检验的步骤

A systematic approach helps avoid mistakes. Here are the six essential steps for any hypothesis test.

系统性的方法有助于避免错误。以下是任何假设检验的六个基本步骤。

Step 1 — State the null and alternative hypotheses clearly.
步骤 1——清楚陈述原假设和备择假设。
Step 2 — Choose the significance level α (usually 0.05).
步骤 2——选择显著性水平 α（通常为 0.05）。
Step 3 — Determine the appropriate test statistic and its distribution.
步骤 3——确定合适的检验统计量及其分布。
Step 4 — Calculate the test statistic from the sample data.
步骤 4——根据样本数据计算检验统计量。
Step 5 — Find the p-value or compare with critical values.
步骤 5——计算 p 值或与临界值比较。
Step 6 — Make a decision and write a conclusion in context.
步骤 6——做出决策并在实际语境中写出结论。

9. Worked Example: z-Test | 例题解析：z 检验

A manufacturer claims that the mean weight of their cereal boxes is 500 g. A consumer group suspects that the boxes are underweight. They weigh 36 randomly selected boxes and find a sample mean of 495 g. The population standard deviation is known to be 18 g. Test at the 1% significance level whether the mean weight is less than 500 g.

某制造商声称他们的麦片盒平均重量为 500 克。一个消费者组织怀疑这些盒子重量不足。他们随机选取了 36 盒进行称重，发现样本均值为 495 克。已知总体标准差为 18 克。在 1% 的显著性水平下检验平均重量是否小于 500 克。

Solution / 解答：

Step 1 — Hypotheses. Since we suspect underweight, use a left-tailed test:

步骤 1 — 假设。由于怀疑重量不足，使用左尾检验：

H₀: μ = 500 vs H₁: μ < 500

Step 2 — Significance level: α = 0.01.

步骤 2 — 显著性水平：α = 0.01。

Step 3 — Test statistic. Since σ is known, use the z-statistic:

步骤 3 — 检验统计量。由于 σ 已知，使用 z 统计量：

z = (495 − 500) / (18 / √36) = −5 / 3 = −1.667

Step 4 — Critical value. For α = 0.01 left-tailed, the critical z is −2.326. The p-value is P(Z < −1.667) = 0.0478.

步骤 4 — 临界值。对于 α = 0.01 的左尾检验，临界 z 值为 −2.326。p 值为 P(Z < −1.667) = 0.0478。

Step 5 — Decision. Since −1.667 > −2.326, and p-value 0.0478 > 0.01, we fail to reject H₀.

步骤 5 — 决策。由于 −1.667 > −2.326，且 p 值 0.0478 > 0.01，我们不能拒绝 H₀。

Step 6 — Conclusion. There is insufficient evidence at the 1% significance level to conclude that the boxes are underweight.

步骤 6 — 结论。在 1% 的显著性水平下，没有足够的证据表明盒子重量不足。

10. Worked Example: t-Test | 例题解析：t 检验

A teacher believes that the average score on a test is not equal to 70. A sample of 10 students gives a mean of 74 and a sample standard deviation of 6. Test at the 5% significance level.

一位老师认为某次测试的平均分数不等于 70。10 名学生的样本均值为 74，样本标准差为 6。在 5% 的显著性水平下进行检验。

Solution / 解答：

Step 1 — Hypotheses. This is a two-tailed test:

步骤 1 — 假设。这是一个双尾检验：

H₀: μ = 70 vs H₁: μ ≠ 70

Step 2 — Significance level: α = 0.05.

步骤 2 — 显著性水平：α = 0.05。

Step 3 — Since σ is unknown and n = 10, use the t-statistic with 9 degrees of freedom:

步骤 3 — 由于 σ 未知且 n = 10，使用自由度为 9 的 t 统计量：

t = (74 − 70) / (6 / √10) = 4 / 1.897 = 2.108

Step 4 — Critical value. For a two-tailed t-test at α = 0.05 with df = 9, the critical values are ±2.262.

步骤 4 — 临界值。对于自由度为 9、α = 0.05 的双尾 t 检验，临界值为 ±2.262。

Step 5 — Decision. Since 2.108 lies between −2.262 and 2.262, we fail to reject H₀.

步骤 5 — 决策。由于 2.108 位于 −2.262 和 2.262 之间，我们不能拒绝 H₀。

Step 6 — Conclusion. There is insufficient evidence at the 5% level to conclude that the average score differs from 70.

步骤 6 — 结论。在 5% 的显著性水平下，没有足够的证据表明平均分数不同于 70。

11. Common Misconceptions | 常见误区

Even advanced students often misunderstand certain aspects of hypothesis testing. Let us clarify a few key points.

即使是高年级学生，也常常误解假设检验的某些方面。让我们澄清几个关键点。

“Failing to reject H₀” does not mean H₀ is true. It simply means there is not enough evidence against it.
“不能拒绝 H₀” 并不意味着 H₀ 为真。这只是说没有足够的证据反对它。
The p-value is not the probability that H₀ is true. It is the probability of observing data as extreme as ours, given that H₀ is true.
p 值不是 H₀ 为真的概率。它是在 H₀ 为真的前提下，观测到与当前数据一样极端的结果的概率。
A smaller α does not always mean a better test. Decreasing α reduces Type I error but increases Type II error, reducing power.
α 越小并不总是意味着检验越好。降低 α 会减少第一类错误，但会增加第二类错误，从而降低检验功效。
Statistical significance is not practical significance. A tiny effect can be statistically significant with a huge sample size.
统计显著不等于实际显著。当样本量非常大时，一个微小的效应也可能在统计上显著。

12. Choosing the Correct Test | 选择合适的检验

Selecting the right test depends on the data type, the sample size, and whether the population variance is known.

选择合适的检验取决于数据类型、样本量以及总体方差是否已知。

Situation	Test to Use	Conditions
Test about a mean, σ known	z-test	n large (n ≥ 30) or population is normal
Test about a mean, σ unknown	t-test	Sample from normal distribution; df = n − 1
Test about a proportion	z-test for proportion	np ≥ 10 and n(1−p) ≥ 10
Compare two means	Two-sample t-test	Independent samples

In IB exams, always check whether the population standard deviation is given. If it is given, use the z-test. If only the sample standard deviation is available, use the t-test when the sample size is small.

在 IB 考试中，务必检查是否给出了总体标准差。如果给出了，使用 z 检验。如果只有样本标准差，且样本量较小时，使用 t 检验。

Mastering hypothesis testing requires practice and careful attention to notation, tails, and conclusions. Always express your final answer in the context of the problem.

掌握假设检验需要反复练习，并仔细注意符号、尾部和结论的表达。始终在问题的实际语境中写出最终答案。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

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