📚 Mean of a normal distribution with unknown variance | 方差未知时正态分布均值的推断
In statistical inference, we often need to estimate or test the population mean μ of a normally distributed variable when the population variance σ² is unknown. This scenario occurs frequently because in real-world data collection we rarely know the true variability of the process. Instead of using the standard normal (z) procedures, we rely on the Student's t-distribution, which accounts for the extra uncertainty introduced by estimating σ with the sample standard deviation s. This article covers the essential theory and methods required for Edexcel A-Level Mathematics, including the t-statistic, confidence intervals, hypothesis testing, and worked examples.
在统计推断中，当总体方差 σ² 未知时，我们经常需要估计或检验正态分布总体的均值 μ。这种情况在实际数据收集中非常普遍，因为我们很少真正了解过程的真实变异性。此时不再使用基于标准正态分布的 z 方法，而是借助学生氏 t 分布，该分布能够反映用样本标准差 s 估计 σ 所引入的额外不确定性。本文涵盖 Edexcel A-Level 数学所需的核心理论和方法，包括 t 统计量、置信区间、假设检验以及计算示例。
1. Introduction to Inference for a Mean with Unknown Variance | 未知方差下均值推断简介
When a population is normally distributed with unknown variance σ², the sample mean x̄ remains the best point estimate of μ. However, the standard error of x̄, which is σ/√n, cannot be computed directly because σ is unknown. We must replace σ with the sample standard deviation s, resulting in the estimated standard error s/√n. This substitution changes the sampling distribution of the standardized mean from a standard normal distribution to a t-distribution with n−1 degrees of freedom, provided the sample is drawn from a normal population.
当总体服从正态分布但方差 σ² 未知时，样本均值 x̄ 仍然是 μ 的最佳点估计。但由于 σ 未知，x̄ 的标准误差 σ/√n 无法直接计算。我们必须用样本标准差 s 代替 σ，得到估计的标准误差 s/√n。这一替换将标准化均值的抽样分布从标准正态分布变为自由度为 n−1 的 t 分布，前提是样本来自正态总体。
Understanding this shift is crucial for correct confidence intervals and hypothesis tests. The t-distribution is slightly wider and has heavier tails than the normal distribution, which compensates for the variability in s. As the sample size increases, the t-distribution approaches the normal distribution, making the distinction less important for large n.
理解这一转变对于构建正确的置信区间和进行假设检验至关重要。t 分布比正态分布略宽且尾部更厚，这补偿了 s 本身的不确定性。随着样本量增大，t 分布趋近于正态分布，因此当 n 很大时，两者的区别变得不那么重要。
2. The Sample Mean and Its Distribution | 样本均值及其分布
Let X₁, X₂, …, Xₙ be a random sample from a normal distribution N(μ, σ²). The sample mean is defined as:
x̄ = (1/n) Σ Xᵢ
The sampling distribution of x̄ is also normal:
x̄ ~ N(μ, σ²/n)
设 X₁, X₂, …, Xₙ 是从正态总体 N(μ, σ²) 中抽取的随机样本。样本均值定义为：
x̄ = (1/n) Σ Xᵢ
x̄ 的抽样分布也是正态分布：
x̄ ~ N(μ, σ²/n)
This result holds regardless of the sample size because the population is normal. If σ were known, the standardized statistic (x̄ − μ) / (σ/√n) would follow a standard normal distribution N(0,1). However, when σ is replaced by s, we must use a t-distribution to account for the extra variation.
由于总体服从正态分布，无论样本容量大小，这一结论都成立。如果 σ 已知，标准化统计量 (x̄ − μ) / (σ/√n) 服从标准正态分布 N(0,1)。但当 σ 被 s 替代时，我们必须使用 t 分布以反映额外的变异。
3. Estimating Population Variance: Sample Variance | 估计总体方差：样本方差
The unknown population variance σ² is estimated by the sample variance s², computed as:
s² = Σ (Xᵢ − x̄)² / (n − 1)
1 instead of n to obtain an unbiased estimator of σ². The sample standard deviation is s = √s². Because s varies from sample to sample, the denominator of the test statistic becomes a random variable, requiring a distribution that has fatter tails than the normal—this is the Student's t-distribution.
未知的总体方差 σ² 由样本方差 s² 进行估计，计算公式为：
s² = Σ (Xᵢ − x̄)² / (n − 1)
我们除以 n−1 而非 n，是为了获得 σ² 的无偏估计量。样本标准差为 s = √s²。由于 s 随样本变化，检验统计量的分母成为随机变量，这就要求使用一种尾部比正态分布更厚的分布——这就是学生氏 t 分布。
4. The t-Statistic | t 统计量
To test hypotheses or construct intervals for μ when σ is unknown, we use the t-statistic:
t = (x̄ − μ₀) / (s / √n)
s statistic follows a t-distribution with ν = n−1 degrees of freedom. The t-statistic measures how many estimated standard errors the sample mean lies from the hypothesised mean.
当 σ 未知时，检验 μ 或构建其置信区间需使用 t 统计量：
t = (x̄ − μ₀) / (s / √n)
其中 μ₀ 是总体均值的假设值。若原假设 H₀: μ = μ₀ 为真且样本来自正态总体，该统计量服从自由度为 ν = n−1 的 t 分布。t 统计量度量了样本均值与假设均值之间相差多少个估计的标准误差。
5. The Student's t-Distribution | 学生氏 t 分布
The Student's t-distribution is a family of symmetric, bell-shaped curves that are similar to the standard normal but with heavier tails. Its shape depends on the degrees of freedom (df). As df increases, the t-distribution approaches N(0,1). At lower df (small samples), the tails are significantly fatter, reflecting greater uncertainty about σ.
学生氏 t 分布是一族对称的钟形曲线，与标准正态分布相似但尾部更厚。其形状取决于自由度 (df)。随着自由度增加，t 分布趋近于 N(0,1)。在自由度较低（小样本）时，尾部明显更厚，反映出对 σ 的估计具有更大的不确定性。
For a given significance level, the critical values from the t-distribution are larger in magnitude than the corresponding z-values. This means confidence intervals will be wider and hypothesis tests more conservative, preventing an inflation of type I error due to estimating σ.
对于给定的显著性水平，t 分布的临界值在绝对值上大于相应的 z 值。这意味着置信区间将更宽，假设检验也更加保守，从而避免了因估计 σ 而导致的第一类错误概率膨胀。
6. Degrees of Freedom | 自由度
ses because we use the data to estimate one parameter (the mean) before computing s². Since the sum of deviations Σ (Xᵢ − x̄ ion of s². The df parameter controls the shape of the t-distribution: the smaller the df, the more spread out the distribution.
单样本 t 检验的自由度为 ν = n−1。这个数目的由来在于，我们在计算 s² 之前已经使用数据估计了一个参数（均值）。由于偏差和 Σ (Xᵢ − x̄) = 0，在 s² 的计算中仅有 n−1 个独立的信息量。自由度参数控制 t 分布的形状：自由度越小，分布越分散。
ch closer to normal. In examination questions, you will typically be given a table of critical t-values or be expected to use a calculator to find them for specified df and tail probabilities.
例如，当 n = 5 时，df = 4，t 分布的尾部非常厚；而当 n = 30，df = 29 时，分布非常接近正态。在考试题目中，通常会提供 t 临界值表，或者要求考生使用计算器查找指定自由度与尾部概率下的临界值。
7. Confidence Intervals for the Mean (σ Unknown) | 均值置信区间（σ未知）
A 100(1−α)% confidence interval for μ when σ is unknown is given by:
x̄ ± tν, α/2 × (s / √n)
where tν, α/2 ees of freedom that puts α/2 in each tail. This interval provides a range of plausible values for μ based on the sample data.
当 σ 未知时，μ 的 100(1−α)% 置信区间为：
x̄ ± tν, α/2 × (s / √n)
其中 tν, α/2 是自由度为 ν = n−1 的 t 分布的双尾临界值，每侧尾部概率为 α/2。该区间根据样本数据给出了 μ 的一组合理取值范围。
To construct a 95% confidence interval with n = 10, you would find t9, 0.025 (often 2.262) and compute the margin of error. The wider interval relative to the z-interval acknowledges the added uncertainty due to unknown σ.
若要构建 n = 10 的 95% 置信区间，需查出 t9, 0.025（通常为 2.262），并计算误差边际。与 z 区间相比，此区间更宽，这体现了未知 σ 带来的额外不确定性。
8. Hypothesis Testing for the Mean (σ Unknown) | 均值假设检验（σ未知）
The steps for a one-sample t-test are:
- sis H₀: μ = μ₀ and the alternative H₁: μ ≠ μ₀ (two-tailed), μ < μ₀ or μ > μ₀ (one-tailed).
- Compute the test statistic t = (x̄ − μ₀) / (s/√n).
- Determine the critical value(s) tν, α/2 or tν, α for the chosen significance level α, with ν = n−1.
- Compare |t| to the critical value, or use the p-value approach, and draw a conclusion in context.
单样本 t 检验的步骤为：
- 陈述原假设 H₀: μ = μ₀ 和备择假设 H₁: μ ≠ μ₀（双尾）、μ < μ₀ 或 μ > μ₀（单尾）。
- 计算检验统计量 t = (x̄ − μ₀) / (s/√n)。
- 根据选定的显著性水平 α，查得临界值 tν, α/2 或 tν, α，自由度 ν = n−1。
- 将 |t| 与临界值比较，或使用 p 值法，并结合实际背景得出结论。
It is essential to check the normality assumption. The t-test is robust to mild departures from normality, especially for larger sample sizes, but for very small samples the data should be approximately normal. The test assumes independent observations and a continuous variable.
检验的正态性假定至关重要。t 检验对于轻微偏离正态具有一定稳健性，尤其是在样本量较大时，但对于非常小的样本，数据应近似服从正态分布。检验假定观测值独立且变量连续。
9. Worked Example: Confidence Interval | 计算示例：置信区间
A random sample of 8 observations from a normal population yields: x̄ = 12.4, s = 2.1. Find a 95% confidence interval for the population mean μ.
Solution: n = 8, df = 7. For 95% confidence, t7, 0.025 = 2.365 (from tables). The margin of error is 2.365 × (2.1/√8 1.756. The confidence interval is (12.4 − 1.756, 12.4 + 1.756) = (10.64, 14.16). We are 95% confident that the true mean lies in this interval.
从一个正态总体中抽取容量为 8 的随机样本，得到：x̄ = 12.4，s = 2.1。求总体均值 μ 的 95% 置信区间。
解答：n = 8，df = 7。95% 置信水平下，t7, 0.025 = 2.365（查表）。误差边际为 2.365 × (2.1/√8) = 2.365 × 0.7425 ≈ 1.756。置信区间为 (12.4 − 1.756, 12.4 + 1.756) = (10.64, 14.16)。我们有 95% 的把握认为总体均值落在该区间内。
10. Worked Example: Hypothesis Test | 计算示例：假设检验
er claims the mean lifespan of a component is 500 hours. A sample of 10 components gives x̄ = 483 hours, s = 22 hours. Test at the 5% significance level whether the true mean is less than 500 hours.
H₀: μ = 500, H₁: μ < 500 (left-tailed test). Test statistic: t = (483 − 500) / (22/√10) = −17 / 6.957 = −2.444. Degrees of freedom ν = 9. One-tailed critical value t9, 0.05 egion is t < −1.833). Since −2.444 < −1.833, we reject H₀. There is sufficient evidence to conclude the mean lifespan is less than 500 hours.
某制造商声称一个组件的平均寿命为 500 小时。一个容量为 10 的样本得出 x̄ = 483 小时，s = 22 小时。在 5% 的显著性水平下，检验真实均值是否小于 500 小时。
H₀: μ = 500，H₁: μ < 500（左尾检验）。检验统计量：t = (483 − 500) / (22/√10) = −17 / 6.957 = −2.444。自由度 ν = 9。单尾临界值 t9, 0.05 = 1.833（由对称性，拒绝域为 t < −1.833）。由于 −2.444 < −1.833，拒绝 H₀。有充分证据表明平均寿命低于 500 小时。
11. Comparison with the z-Test and When to Use t | 与 z 检验的比较及使用时机
The table below summarises the main differences between the z and t procedures for the mean of a normal distribution.
| Feature | z-test (σ known) | t-test (σ unknown) |
|---|---|---|
| Test statistic | z = (x̄ − μ₀) / (σ/√n) | t = (x̄ − μ₀) / (s/√n) |
| Distribution | Standard normal N(0,1) | t-distribution, df = n−1 |
| Critical values | Fixed (e.g., 1.96 for 95%) | Depend on df; larger for small n |
| Sample size requirement | Central Limit Theorem allows large n if not normal | Assumes normality; robust for moderate n |
下表总结了正态分布均值推断中 z 方法与 t 方法的主要区别。
| 特征 | z 检验 (σ 已知) | t 检验 (σ 未知) |
|---|---|---|
| 检验统计量 | z = (x̄ − μ₀) / (σ/√n) | t = (x̄ − μ₀) / (s/√n) |
| 分布 | 标准正态 N(0,1) | t 分布，df = n−1 |
| 临界值 | 固定（如 95% 为 1.96） | 依赖于 df；小样本时更大 |
| 样本量要求 | 若非正态，中心极限定理允许大样本 | 假定正态；中等样本时稳健 |
Use the t-test whenever the population standard deviation is unknown and must be estimated from the sample. Even when σ is known, if the sample size is small and normality cannot be guaranteed, the t-test may still be preferred for its conservatism. For very large samples (n > 100), the t and z critical values are nearly identical, so the distinction becomes negligible in practice.
只要总体标准差未知且必须从样本中估计，就应使用 t 检验。即使 σ 已知，若样本量小且无法保证正态性，出于保守考虑也可能更偏爱 t 检验。对于非常大的样本（n > 100），t 与 z 的临界值近乎相同，因此实际应用中区别可忽略。
12. Summary and Key Points | 总结与要点
In A-Level Mathematics, the topic 'Mean of a normal distribution with unknown variance' equips you with the tools to handle realistic inferential problems. Always remember:
- The t-statistic replaces z when σ is unknown.
- The degrees of freedom for a one-sample t-procedure are n−1.
- Confidence intervals take the form x̄ ± tν, α/2 × s/√n.
- Hypothesis tests compare the observed t with critical values from the t-distribution.
- The normality assumption should be checked, especially for small samples.
在 A-Level 数学中，“方差未知时正态分布均值的推断”这一主题为您提供了处理实际推断问题的工具。务必牢记：
- 当 σ 未知时，用 t 统计量代替 z。
- 单样本 t 过程的自由度为 n−1。
- 置信区间的形式为 x̄ ± tν, α/2 × s/√n。
- 假设检验将观察到的 t 值与 t 分布的临界值进行比较。
- 正态性假定需要检验，尤其对于小样本而言。
Mastery of these concepts not only aids exam success but also builds a solid foundation for further study in statistics and data analysis.
掌握这些概念不仅有助于考试成功，也为进一步学习统计学和数据分析奠定坚实基础。
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