Unbiased Estimators and Confidence Intervals | 无偏估计量与置信区间

📚 Unbiased Estimators and Confidence Intervals | 无偏估计量与置信区间

In statistical inference, we use sample data to draw conclusions about population parameters. Two foundational concepts — unbiased estimators and confidence intervals — allow us to make reliable estimates and quantify the uncertainty inherent in sampling. This article covers both concepts in depth, tailored for the CIE IB Mathematics curriculum.

在统计推断中，我们利用样本数据来对总体参数得出结论。无偏估计量和置信区间这两个基础概念，使我们能够做出可靠的估计，并对抽样中固有的不确定性进行量化。本文围绕 CIE IB 数学考纲，深入讲解这两个概念。

1. What is an Estimator? | 什么是估计量

An estimator is a rule or formula that uses sample data to estimate an unknown population parameter. For example, the sample mean is an estimator for the population mean μ. The value produced by an estimator from a particular sample is called an estimate.

估计量是利用样本数据估计未知总体参数的规则或公式。例如，样本均值就是总体均值 μ 的一个估计量。由某个特定样本所计算出的具体数值称为估计值。

Parameter: A numerical characteristic of a population, such as μ or σ².

参数：总体的数值特征，如 μ 或 σ²。
Statistic: A numerical characteristic of a sample, such as x̄ or s².

统计量：样本的数值特征，如 x̄ 或 s²。
Point estimate: A single value used to approximate a parameter.

点估计：用于近似总体参数的单一数值。

2. Sampling Distributions | 抽样分布

A sampling distribution is the probability distribution of a statistic obtained from a large number of samples drawn from a specific population. Understanding this distribution is essential for evaluating the quality of an estimator.

抽样分布是指从特定总体中抽取大量样本后，某个统计量的概率分布。理解抽样分布是评估估计量质量的关键。

For a sample of size n from a population with mean μ and variance σ²: E(X̄) = μ, Var(X̄) = σ²⁄n

对于来自均值为 μ、方差为 σ² 的总体的容量为 n 的样本：E(X̄) = μ，Var(X̄) = σ²⁄n

The standard deviation of the sampling distribution, σ/√n, is called the standard error of the mean.

抽样分布的标准差 σ/√n 称为均值的标准误。

3. Unbiased Estimators | 无偏估计量

An estimator is unbiased if its expected value equals the true population parameter. In other words, on average, across many samples, the estimator hits the target parameter exactly.

如果一个估计量的期望值等于真实的总体参数，则该估计量是无偏的。换言之，在大量样本中取平均后，该估计量恰好命中目标参数。

E(θ̂) = θ ⟺ θ̂ is an unbiased estimator of θ

E(θ̂) = θ ⟺ θ̂ 是 θ 的无偏估计量

If E(θ̂) ≠ θ, the estimator is said to be biased. The bias is defined as B(θ̂) = E(θ̂) − θ.

如果 E(θ̂) ≠ θ，则称该估计量有偏。偏差定义为 B(θ̂) = E(θ̂) − θ。

4. Common Unbiased Estimators | 常见的无偏估计量

In the IB syllabus, two key unbiased estimators are emphasized:

在 IB 课程大纲中，重点强调两个关键的无偏估计量：

Parameter \| 参数	Unbiased Estimator \| 无偏估计量
Population mean μ 总体均值 μ	Sample mean X̄ = (1/n)ΣXᵢ 样本均值 X̄ = (1/n)ΣXᵢ
Population variance σ² 总体方差 σ²	Sample variance S² = (1/(n−1))Σ(Xᵢ − X̄)² 样本方差 S² = (1/(n−1))Σ(Xᵢ − X̄)²

Notice the divisor n − 1 rather than n in the sample variance. This correction, known as Bessel’s correction, ensures unbiasedness.

注意样本方差中除以 n − 1 而不是 n。这一修正称为贝塞尔修正，它保证了无偏性。

5. Why n − 1 Instead of n? | 为什么用 n − 1 而不是 n

Intuitively, the sum of deviations Σ(Xᵢ − X̄) always equals zero, so only n − 1 of the deviations are independent. Using n would systematically underestimate the true variance.

直观地看，Σ(Xᵢ − X̄) 总是等于零，因此 n 个偏差中只有 n − 1 个是独立的。使用 n 会系统性低估真实的方差。

E[(1/n)Σ(Xᵢ − X̄)²] = ((n−1)/n)σ² ≠ σ²

Therefore, we multiply by n/(n−1) to fix the bias, producing the unbiased estimator S².

因此，我们乘以 n/(n−1) 来修正偏差，从而得到无偏估计量 S²。

6. Standard Error | 标准误

The standard error (SE) measures the precision of an estimator. For the sample mean:

标准误（SE）衡量估计量的精度。对于样本均值：

SE(X̄) = σ⁄√n (when σ is known) | SE(X̄) = S⁄√n (when σ is unknown)

SE(X̄) = σ⁄√n（σ 已知时） | SE(X̄) = S⁄√n（σ 未知时）

A smaller standard error indicates a more precise estimator. Notice that SE decreases as sample size n increases — larger samples yield sharper estimates.

标准误越小表明估计量越精确。注意 SE 随着样本量 n 的增大而减小——大样本产生更精确的估计。

7. Introduction to Confidence Intervals | 置信区间入门

A confidence interval (CI) is a range of values, calculated from sample data, that is likely to contain the true population parameter. The confidence level, typically 95% or 99%, reflects the long-run proportion of intervals that would capture the parameter if we repeated sampling many times.

置信区间（CI）是根据样本数据计算出的一个数值范围，该范围有可能包含真实的总体参数。置信水平（通常为 95% 或 99%）反映了若我们重复抽样多次，能够包含真实参数的区间所占的长期比例。

A 95% CI for μ: X̄ ± 1.96 × σ⁄√n

μ 的 95% 置信区间：X̄ ± 1.96 × σ⁄√n

The multiplier 1.96 is the z-critical value for a 95% confidence level. Common z-values include 1.645 (90%), 1.96 (95%), and 2.576 (99%).

乘数 1.96 是 95% 置信水平对应的 z 临界值。常用 z 值包括 1.645（90%）、1.96（95%）和 2.576（99%）。

8. Confidence Interval when σ is Unknown | σ 未知时的置信区间

In most real situations, the population variance σ² is unknown. We then replace σ with the sample standard deviation S and use the t-distribution with n − 1 degrees of freedom:

在大多数实际情况中，总体方差 σ² 是未知的。此时我们用样本标准差 S 替代 σ，并使用自由度为 n − 1 的 t 分布：

A 95% CI: X̄ ± t₀.₀₂₅(n−1) × S⁄√n

95% 置信区间：X̄ ± t₀.₀₂₅(n−1) × S⁄√n

For large samples (n ≥ 30), the t-distribution closely approximates the standard normal, so z-values may be used as a reasonable approximation.

对于大样本（n ≥ 30），t 分布非常接近标准正态分布，因此可以使用 z 值作为合理的近似。

9. Interpreting Confidence Intervals | 置信区间的解读

A common misconception is that a 95% CI means there is a 95% probability the parameter lies in the interval. In the frequentist framework, the parameter is fixed, and the interval is random.

一个常见的误解是 95% 置信区间意味着参数落在区间内的概率为 95%。在频率学框架下，参数是固定的，区间才是随机的。

Correct interpretation: If we repeatedly took samples and constructed confidence intervals in the same way, approximately 95% of those intervals would contain the true parameter μ.

正确的解读：如果我们反复抽取样本并以相同方式构建置信区间，那么大约 95% 的此类区间将包含真实的总体参数 μ。

Incorrect interpretation: There is a 95% chance that μ lies between L and U for a specific interval.

错误的解读：对于某一个具体区间，μ落在其间的概率为 95%。

10. Worked Example | 例题解析

Problem: A random sample of 40 students shows a mean score of 72 with a sample standard deviation of 8. Construct a 95% confidence interval for the population mean score.

题目：从某学校随机抽取 40 名学生，其平均分为 72，样本标准差为 8。求总体平均分的 95% 置信区间。

Solution: Since n = 40 ≥ 30, use the z-approach:

解答：由于 n = 40 ≥ 30，采用 z 方法：

SE = 8/√40 = 1.265; Margin of error = 1.96 × 1.265 = 2.479

SE = 8/√40 = 1.265；误差幅度 = 1.96 × 1.265 = 2.479

CI = 72 ± 2.479 = (69.52, 74.48)

置信区间 = 72 ± 2.479 = (69.52, 74.48)

We are 95% confident that the true population mean lies between 69.52 and 74.48.

我们有 95% 的把握认为总体真实均值介于 69.52 与 74.48 之间。

11. Factors Affecting Interval Width | 影响区间宽度的因素

Three factors determine the width of a confidence interval:

三个因素决定置信区间的宽度：

Sample size n: Larger n produces narrower intervals (SE decreases as 1/√n).

样本量 n：n 越大，区间越窄（SE 按 1/√n 减小）。
Confidence level: Higher confidence requires a larger critical value, widening the interval.

置信水平：置信度越高，临界值越大，区间越宽。
Population variability σ: More variable populations yield wider intervals.

总体变异度 σ：总体变异越大，区间越宽。

Factor \| 因素	Increase \| 增加	Effect on Width \| 对宽度的影响
n 样本量	↑	Decreases 变窄
Confidence level 置信水平	↑	Increases 变宽
σ 总体标准差	↑	Increases 变宽

12. Common Pitfalls and Exam Tips | 常见易错点与考试技巧

Students frequently make the following mistakes:

同学们经常犯以下错误：

Using n instead of n − 1 when computing sample variance. Always use n − 1 for S².

计算样本方差时使用 n 而非 n − 1。计算 S² 时务必使用 n − 1。
Misinterpreting the confidence level as the probability that μ lies in a specific interval. Instead, it describes the method’s reliability.

误解读置信水平为 μ 落在某个具体区间内的概率。实际上它描述的是该方法的可靠性。
Confusing S and σ in the standard error formula. Use σ only when it is given; otherwise use S.

混淆 S 与 σ在标准误公式中的使用。只有题目给出 σ 时才能使用；否则用 S。
Rounding too early in calculations — keep at least 4 significant figures until the final answer.

过早四舍五入——在得到最终答案前至少保留 4 位有效数字。

In CIE exams, carefully note whether σ is known, and always state the distribution you are using (z or t). Clearly present the formula before substituting values.

在 CIE 考试中，仔细判断 σ 是否已知，并明确说明使用的是 z 分布还是 t 分布。先写出公式，再代入数值，步骤要清晰规范。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Unbiased Estimators and Confidence Intervals | 无偏估计量与置信区间