📚 Confidence Interval for the Mean with Unknown Variance (from Data) | 未知方差下均值的置信区间（基于数据）
In inferential statistics, constructing a confidence interval for a population mean μ is a fundamental technique. When the population standard deviation σ is known, the normal distribution provides the critical values. In practice, however, σ is rarely known and must be estimated from sample data using the sample standard deviation s. This extra layer of estimation leads to a wider, more realistic interval based on Student's t‑distribution. In this article, we explore how to compute and interpret a confidence interval for μ when the population variance is unknown, using only a given data set.
在推断统计中，构建总体均值 μ 的置信区间是一项基本技术。当总体标准差 σ 已知时，使用正态分布提供临界值。然而在实践中 σ 几乎总是未知的，必须用样本标准差 s 从样本数据中估计。这一层额外的估计导致了一个更宽、更符合实际的区间，它基于学生 t 分布。本文将探讨在总体方差未知时，如何仅利用给定数据集计算并解释 μ 的置信区间。
1. Why the t‑Distribution? | 为什么使用 t 分布？
When σ is unknown, we replace it with the sample standard deviation s. The statistic (x̄ − μ)/(s / √n) no longer follows a standard normal distribution exactly. Instead, it follows a t‑distribution with n−1 degrees of freedom, provided the population is normally distributed or the sample size is large enough.
当 σ 未知时，我们用样本标准差 s 替代它。统计量 (x̄ − μ)/(s / √n) 不再精确服从标准正态分布，而是服从自由度为 n−1 的 t 分布，前提是总体服从正态分布或样本量足够大。
The shape of the t‑distribution accounts for the extra uncertainty introduced by estimating σ with s. This uncertainty is greater for small samples and gradually disappears as n increases.
t 分布的形状反映了用 s 估计 σ 所引入的额外不确定性。这种不确定性在小样本下更大，并随着 n 的增加而逐渐消失。
2. Properties of the t‑Distribution | t 分布的性质
The t‑distribution is symmetric and bell‑ er ν = n − 1 controls the tail thickness: lower ν gives fatter tails, while as ν → ∞ the t‑distribution approaches the standard normal distribution.
t 分布与标准正态分布一样对称且呈钟形，但尾部更厚。自由度参数 ν = n − 1 控制尾部厚度：ν 越小尾部越厚，当 ν → ∞ 时 t 分布趋近于标准正态分布。
Critical values tα/2, ν are larger than the corresponding zα/2 for the same confidence level, especially when ν is small. This increases the margin of error and produces a wider interval, compensating for the use of s in place of σ.
在相同置信水平下，临界值 tα/2, ν 大于对应的 zα/2，尤其当 ν 较小时。这增大了误差范围，使区间更宽，弥补了用 s 替代 σ 所带来的不确定性。
3. The Confidence Interval Formula | 置信区间公式
The 100(1−α)% confidence interval for μ when σ is unknown is given by:
当 σ 未知时，μ 的 100(1−α)% 置信区间由下式给出：
x̄ ± tα/2, n−1 × (s / √n)
Here x̄ is the sample mean, s is the sample standard deviation, n is the sample size, and tα/2, n−1 is the critical value from the t‑distribution with n−1 degrees of freedom, cutting off an area of α/2 in each tail.
这里 x̄ 是样本均值，s 是样本标准差，n 是样本容量，tα/2, n−1 是自由度为 n−1 的 t 分布临界值，其在每一尾部截断面积为 α/2。
The quantity s / √n is called the estimated standard error of the mean. The margin of error is E = tα/2, n−1 × (s / √n). The interval is then (x̄ − E, x̄ + E).
量 s / √n 称为均值的估计标准误。误差范围 E = tα/2, n−1 × (s / √n)，区间为 (x̄ − E, x̄ + E)。
4. Assumptions and Conditions | 假设与条件
To use the t‑interval reliably, the following conditions should be met:
为了可靠地使用 t 区间，应满足以下条件：
- Random sample: the data must be obtained through a random process so that the sample is representative.
- 随机样本：数据必须通过随机过程获得，使样本具有代表性。
- Independence: observations should be independent of each other (e.g. the sample size is less than 10% of the population when sampling without replacement).
- 独立性：观测值之间应相互独立（例如，在不放回抽样时样本量小于总体的 10%）。
- Approximate normality: the population from which the sample is drawn should be normally distributed, especially for small n (n < 30). For larger samples, the Central Limit Theorem ensures that the sampling distribution of x̄ is approximately normal, making the t‑interval robust even if the population is not perfectly normal.
- 近似正态性：样本所来自的总体应服从正态分布，尤其对于小样本（n < 30）。对于较大样本，中心极限定理保证 x̄ 的抽样分布近似正态，因此即使总体不完全正态，t 区间也具有稳健性。
5. Step‑by‑Step Calculation from Data | 基于数据的逐步计算
Given a raw data set, follow these steps to construct the confidence interval:
给定原始数据集，按以下步骤构建置信区间：
1. Compute the sample mean x̄ = (Σxᵢ) / n.
1. 计算样本均值 x̄ = (Σxᵢ) / n。
2. Calculate the sample standard deviation s = √[ Σ(xᵢ − x̄)² / (n − 1) ]. Use the unbiased formula with n−1 in the denominator.
2. 计算样本标准差 s = √[ Σ(xᵢ − x̄)² / (n − 1) ]，使用分母为 n−1 的无偏公式。
3. Determine the degrees of freedom ν = n − 1.
3. 确定自由度 ν = n − 1。
4. Choose the confidence level (1−α) and find the critical t‑value tα/2, ν from a t‑table or technology.
4. 选择置信水平 (1−α)，并从 t 表或技术工具中找到临界值 tα/2, ν。
5. Compute the estimated standard error SE = s / √n.
5. 计算估计标准误 SE = s / √n。
6. Find the margin of error E = tα/2, ν × SE.
6. 计算误差范围 E = tα/2, ν × SE。
7. Construct the interval: (x̄ − E, x̄ + E).
7. 构建区间：(x̄ − E, x̄ + E)。
6. Worked Example | 计算实例
A student records the response times (in milliseconds) of a visual stimulus in a psychology experiment: 250, 245, 255, 260, 248, 252, 257, 253, 249, 246. Construct a 95% confidence interval for the population mean response time.
某学生在心理学实验中记录了对视觉刺激的反应时间（毫秒）：250, 245, 255, 260, 248, 252, 257, 253, 249, 246。构建总体平均反应时间的 95% 置信区间。
First, summarise the data: n = 10, Σx = 2515, x̄ = 251.5 ms.
首先汇总数据：n = 10, Σx = 2515, x̄ = 251.5 ms。
Compute deviations and squared deviations:
计算偏差与平方偏差：
| xᵢ | xᵢ − x̄ | (xᵢ − x̄)² |
| 250 | −1.5 | 2.25 |
| 245 | −6.5 | 42.25 |
| 255 | 3.5 | 12.25 |
| 260 | 8.5 | 72.25 |
| 248 | −3.5 | 12.25 |
| 252 | 0.5 | 0.25 |
| 257 | 5.5 | 30.25 |
| 253 | 1.5 | 2.25 |
| 249 | −2.5 | 6.25 |
| 246 | −5.5 | 30.25 |
Sum of squared deviations = 210.5. Then s = √210.5 / 9 ≈ √23.3889 ≈ 4.836 ms.
平方偏差和 = 210.5。因此 s = √210.5 / 9 ≈ √23.3889 ≈ 4.836 ms。
Degrees of freedom ν = 9. For a 95% confidence level, α = 0.05, α/2 = 0.025, and t0.025, 9 = 2.262 (from t‑table).
自由度 ν = 9。对于 95% 置信水平，α = 0.05，α/2 = 0.025，查表得 t0.025, 9 = 2.262。
Standard error SE = s / √n = 4.836 / √10 ≈ 1.529 ms.
标准误 SE = s / √n = 4.836 / √10 ≈ 1.529 ms。
Margin of error E = 2.262 × 1.529 ≈ 3.46 ms.
误差范围 E = 2.262 × 1.529 ≈ 3.46 ms。
95% confidence interval: (251.5 − 3.46, 251.5 + 3.46) = (248.04, 254.96) ms.
95% 置信区间：(251.5 − 3.46, 251.5 + 3.46) = (248.04, 254.96) ms。
7. Interpreting the Confidence Interval | 解释置信区间
A correct interpretation: "We are 95% confident that the true population mean response time lies between 248.04 ms and 254.96 ms." This means that if we were to take many random samples of the same size and compute a 95% confidence interval from each, about 95% of those intervals would contain the true mean μ.
正确的解释是："我们有 95% 的把握认为，真实的总体平均反应时间介于 248.04 ms 和 254.96 ms 之间。" 这意味着如果我们多次抽取相同大小的随机样本并计算各自的 95% 置信区间，大约 95% 的区间会包含真实的总体均值 μ。
It is incorrect to say "The probability that μ lies in this interval is 0.95." The parameter μ is fixed; it is the interval that varies from sample to sample.
错误地说法是："μ 位于此区间的概率为 0.95。" 参数 μ 是固定的，随样本变化的是区间。
8. Effect of Confidence Level | 置信水平的影响
Increasing the confidence level (e.g. from 95% to 99%) raises the critical t‑value, which widens the margin of error and produces a broader interval. Conversely, lowering the confidence level yields a narrower interval but with less assurance of capturing μ.
提高置信水平（例如从 95% 到 99%）会增大临界 t 值，从而加大误差范围，产生更宽的区间。反之，降低置信水平会得到较窄的区间，但捕获 μ 的信心较低。
The choice of confidence level depends on the context: high‑risk scenarios (medical trials) often use 99%, while exploratory analyses may use 90%.
置信水平的选择取决于背景：高风险场景（医学试验）常用 99%，而探索性分析可能使用 90%。
9. Effect of Sample Size | 样本量的影响
For a fixed confidence level and approximately constant s, increasing the sample size n reduces the standard error (s / √n) and decreases the margin of error. Additionally, the t‑critical value shrinks as degrees of freedom increase, gradually approaching the z‑value.
在固定的置信水平和近似恒定的 s 下，增大样本量 n 会减小标准误 (s / √n)，从而减小误差范围。此外，t 临界值随着自由度增加而缩小，逐渐趋近 z 值。
Thus, collecting more data is a direct way to obtain more precise estimates of μ. However, the benefit diminishes because of the √n in the denominator – quadrupling n roughly halves the margin of error.
因此，收集更多数据是获得更精确 μ 估计的直接途径。但这种收益会递减，因为分母中有 √n——将 n 增加为原来的四倍大约能使误差范围减半。
10. Common Pitfalls | 常见误区
A frequent error is to use the z‑distribution when σ is unknown and the sample size is small. This leads to an interval that is too narrow and underestimates uncertainty. Always check whether σ is known; if not, default to the t‑interval.
一个常见错误是在 σ 未知且样本量较小时使用 z 分布。这会导致区间过窄，低估不确定性。务必检查 σ 是否已知；若未知，默认使用 t 区间。
Another mistake is using n instead of n−1 when computing s. The sample standard deviation must divide by n−1 to be an unbiased estimator of σ. This is critical for the t‑interval to maintain its stated confidence level.
另一个错误是在计算 s 时用 n 而不是 n−1 作分母。样本标准差必须除以 n−1，才能成为 σ 的无偏估计量。这对 t 区间维持其所声称的置信水平至关重要。
Some students misinterpret the interval as describing the spread of individual data points rather than the precision of the mean estimate. The confidence interval concerns the location of μ, not the range of future observations.
有些学生将区间解释为单个数据点的散布，而非均值估计的精度。置信区间是关于 μ 的位置，而不是未来观测值的范围。
11. Using Technology: GDC & Tables | 使用技术：图形计算器与表格
In IB examinations, you may be expected to find critical t‑ ility, refer to the table row for df = ν and read the column for the required α/2. On a TI‑84 or similar GDC, use invT( probability, df ) or the built‑in TInterval function for one‑sample data.
在 IB 考试中，你可能需要用统计表或图形计算器（GDC）查找临界 t 值。对于给定的 ν 和尾部概率，在表中找到 df = ν 的行，读取相应 α/2 的列。在 TI‑84 或类似 GDC 上，使用 invT( probability, df ) 或通过单样本数据的 TInterval 功能直接求得区间。
The TInterval command on a GDC not only computes the interval but also displays x̄, s, and the margin of error, which helps in verifying manual calculations. Remember to state the degrees of freedom and the t‑value used when showing working.
GDC 上的 TInterval 命令不仅能计算区间，还显示 x̄、s 和误差范围，有助于检验手工计算。在展示解题过程时，记得写明所用自由度和 t 值。
12. Summary | 总结
ion mean when the variance is unknown is a core skill in IB Mathematics. It requires the t‑ count for the added uncertainty of estimating σ with s. The procedure involves computing x̄ appropriate t critical value with n−1 degrees of freedom, and then applying the formula x̄ ± tα/2, n−1 × (s / √n). Correct interpretation, checking assumptions, and careful use of technology all contribute to a robust statistical conclusion.
在方差未知时构建总体均值置信区间是 IB 数学的核心技能。它需要 t 分布来处理用 s 估计 σ 所增加的不确定性。过程包括计算 x̄ 和 s、确定自由度为 n−1 的恰当 t 临界值，再应用公式 x̄ ± tα/2, n−1 × (s / √n)。正确的解释、检查假设以及谨慎使用技术工具，都有助于得出可靠的统计结论。
Published by TutorHao | IB 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