📚 Estimators, Bias and Standard Error | 估计量、偏差与标准误差
In A-Level Statistics, when we take a sample from a population, we rarely know the true values of the parameters that define that population. The concepts of estimators, bias, and standard error form the foundation for making reliable inferences about population parameters from sample data. An estimator is a rule or formula that tells us how to calculate an estimate from a given sample. Understanding the properties of estimators – particularly whether they are biased and how much they vary – is crucial for deciding which estimator to use and for quantifying the uncertainty in our conclusions. This article unpacks these ideas in the context of the Edexcel A-Level Mathematics specification, using clear definitions, worked-through examples, and a focus on the sample mean and sample variance.
在A-Level统计学中，当我们从总体中抽取样本时，我们很少知道定义该总体的参数的真实值。估计量、偏差与标准误差这些概念构成了从样本数据中对总体参数进行可靠推断的基础。估计量是一条规则或公式，它告诉我们如何从给定的样本中计算出一个估计值。理解估计量的性质——特别是有没有偏差以及它们的波动程度——对于选择使用哪个估计量以及量化我们结论中的不确定性至关重要。本文在Edexcel A-Level数学考试大纲的背景下，通过清晰的定义、实际演算的例子，并重点关注样本均值和样本方差，来剖析这些概念。
1. Populations, Parameters and Samples | 总体、参数与样本
A population is the entire collection of individuals or items that we are interested in studying. A parameter is a numerical characteristic of a population, such as its mean μ (mu) or its variance σ² (sigma squared). Since dealing with whole populations is often impractical, we draw a random sample of size n, and we use sample statistics to estimate the unknown parameters. A statistic that is used to estimate a population parameter is called an estimator when considered as a rule, and an estimate is the actual numerical value obtained from a particular sample.
总体是我们感兴趣的整个个体或项目的集合。参数是总体的一个数值特征，例如其均值 μ 或方差 σ²。因为处理整个总体往往不可行，所以我们抽取一个容量为 n 的随机样本，并使用样本统计量来估计未知参数。当一个统计量被用作估计总体参数的规则时，它被称为估计量，而从特定样本中得到的实际数值则称为估计值。
For example, if we wish to know the average height of 16-year-old students in the UK, the population mean μ is the parameter. We take a random sample of 50 students, compute the sample mean x̄ (x-bar) using the formula Σx/n, and this x̄ is our estimator. The value we get, say 165.2 cm, is the estimate.
例如，如果我们想知道英国16岁学生的平均身高，总体均值 μ 就是参数。我们随机抽取50名学生，利用公式 Σx/n 计算出样本均值 x̄，这个 x̄ 就是我们的估计量。我们得到的数值，比如165.2 cm，就是估计值。
2. What Is a Point Estimator? | 什么是点估计量？
A point estimator provides a single value as the “best guess” of a population parameter. Common point estimators include the sample mean x̄ for the population mean μ, the sample variance s² for the population variance σ², and the sample proportion p̂ (p-hat) for a population proportion p. In the Edexcel course, you are expected to know the formulas for these estimators and to apply them to both raw data and summary statistics.
点估计量提供一个单一数值作为对总体参数的“最佳猜测”。常见的点估计量包括用样本均值 x̄ 估计总体均值 μ，用样本方差 s² 估计总体方差 σ²，以及用样本比例 p̂ 估计总体比例 p。在Edexcel课程中，你需要掌握这些估计量的公式，并能够将其应用于原始数据和汇总统计量。
The sample mean is defined as:
x̄ = Σxᵢ / n
The sample variance (the unbiased estimator) is defined as:
s² = Σ(xᵢ – x̄)² / (n – 1)
样本均值定义为：x̄ = Σxᵢ / n。样本方差（无偏估计量）定义为：s² = Σ(xᵢ – x̄)² / (n – 1)。
3. Unbiased Estimators | 无偏估计量
e many, many samples of the same size from a population, and calculate the estimator for each sample, the average of all those estimates would be exactly the parameter value. The notation E(θ̂) = θ expresses this property, where θ̂ (theta hat) is the estimator and θ is the parameter.
如果一个估计量的期望值等于真实的总体参数，则称该估计量是无偏的。简而言之，如果我们从总体中抽取很多很多个同样大小的样本，并为每个样本计算出该估计量，那么所有这些估计值的平均值将恰好等于参数值。记号 E(θ̂) = θ 表达了这一性质，其中 θ̂ 是估计量，θ 是参数。
For example, the sample mean x̄ is an unbiased estimator of the population mean μ, because E(X̄ m sample equals the population mean. The proof relies on the linearity of expectation: E(X̄) = E(ΣXᵢ/n) = (1/n)ΣE(Xᵢ) = (1/n) · nμ = μ.
例如，样本均值 x̄ 是总体均值 μ 的无偏估计量，因为 E(X̄) = μ。这是基于这样一个事实：一个随机样本的均值的期望值等于总体均值。证明依赖于期望的线性性质：E(X̄) = E(ΣXᵢ/n) = (1/n)ΣE(Xᵢ) = (1/n) · nμ = μ。
Unbiasedness is a desirable property because it means the estimator does not systematically overestimate or underestimate the parameter. However, it does not guarantee that a single estimate is close to the true value; it only says that the process is correct on average.
无偏性是一个受欢迎的性质，因为它意味着估计量不会系统性地高估或低估参数。然而，它并不能保证单个估计值接近真实值；它只表明这个过程在平均意义下是正确的。
4. Bias of an Estimator | 估计量的偏差
ifference between the expected value of the estimator and the true parameter value: Bias(θ̂) = E(θ̂) – nderestimate. Biased estimators can sometimes be useful if they have other compensating properties like much smaller variance, but at A-Level we focus on recognising bias and correcting it where appropriate.
偏差定义为估计量的期望值与真实参数值之间的差：Bias(θ̂) = E(θ̂) – θ。如果偏差为零，则估计量是无偏的。如果偏差为正，则估计量平均而言倾向于高估 θ；如果为负，则倾向于低估。有时，有偏的估计量如果具有其他补偿性质（比如方差小得多），也可能是有用的，但在A-Level阶段，我们重点在于识别偏差并在适当情况下进行校正。
The most famous example of bias is the “sample variance” when calculated using a divisor of n. If we define σ̂² = Σ(xᵢ – x̄)² / n, then it can be shown that E(σ̂²) = (n – 1)/n · σ², which is less than σ². Thus σ̂ derestimates the population variance on average. That is why the unbiased estimator uses n – 1 in the denominator: s² = Σ(xᵢ – x̄)² / (n – 1).
最著名的偏差例子是使用除数 n 计算“样本方差”时的情况。如果我们定义 σ̂² = Σ(xᵢ – x̄)² / n，那么可以证明 E(σ̂²) = (n – 1)/n · σ²，这小于 σ²。因此 σ̂² 是 σ² 的一个有偏估计量；它平均而言低估了总体方差。这就是为什么无偏估计量在分母中使用 n – 1：s² = Σ(xᵢ – x̄)² / (n – 1)。
Understanding bias is essential when given a formula and asked to determine whether it produces unbiased estimates. You should be comfortable using expectation algebra to find E(θ̂) and comparing it to the parameter.
当给定一个公式并要求判断它是否产生无偏估计时，理解偏差至关重要。你应该能熟练运用期望代数来求出 E(θ̂) 并将其与参数进行比较。
5. Sample Mean as an Unbiased Estimator | 样本均值作为无偏估计量
The sample mean X̄ simple random sample of size n from a population with mean μ and variance σ², we have E(X̄) = μ and Var(X̄ ples produce more precise estimates, but the unbiasedness holds regardless of sample size.
样本均值 X̄ 是估计总体均值 μ 最常用的估计量。它的无偏性是推断的基石。对于一个来自均值为 μ、方差为 σ² 的总体的容量为 n 的简单随机样本，我们有 E(X̄) = μ 且 Var(X̄) = σ²/n。方差公式表明，样本量越大，估计值越精确，但无偏性无论样本大小都成立。
/2 is an unbiased estimator for μ, you simply calculate its expected value: E[(X₁ + X₂)/2] = (μ + μ)/2 = μ. It is unbiased.
这也适用于独立观测值的任何线性组合。例如，如果一道考题要求你验证 (X₁ + X₂)/2 是 μ 的无偏估计量，你只需计算其期望值：E[(X₁ + X₂)/2] = (μ + μ)/2 = μ。它是无偏的。
6. Sample Variance and the n – 1 Correction | 样本方差与 n – 1 校正
When estimating the population variance σ², the statistic S² = Σ(Xᵢ – X̄)² / (n – 1) is the standard unbiased estimator. The use of n – ith n denominator is biased is that the deviations are calculated around the sample mean X̄ tead of the true μ makes the sum of squares slightly smaller on average, so we divide by n – 1 to compensate.
当估计总体方差 σ² 时，统计量 S² = Σ(Xᵢ – X̄)² / (n – 1) 是标准的无偏估计量。分母使用 n – 1 通常称为贝塞尔校正。之所以用 n 作分母的简单平方偏差的均值会有偏，是因为偏离是围绕样本均值 X̄ 计算的，而 X̄ 本身随样本变化。使用样本均值而不是真实的 μ，会使得平方和平均来说略微偏小，因此我们除以 n – 1 来进行补偿。
ow that a given estimator for variance is biased, or to find a constant k such that k Σ(Xᵢ – X̄)² is unbiased. For example, show that for a sample of size 2, the estimator T = (X₁ – X₂)² / 2 is an unbiased estimator of σ². You would compute E(T) = E[(X₁² – ) = σ², you simplify to σ². These exercises reinforce the meaning of bias and expectation.
在Edexcel考试风格的题目中，你可能需要证明一个给定的方差估计量是有偏的，或是找到一个常数 k，使得 k Σ(Xᵢ – X̄)² 是无偏的。例如，证明对于容量为2的样本，估计量 T = (X₁ – X₂)² / 2 是 σ² 的无偏估计量。你会计算出 E(T) = E[(X₁² – 2X₁X₂ + X₂²)/2]，并利用独立性和 E(X₁)=μ、Var(X₁)=σ²，化简后得到 σ²。这些练习强化了对偏差和期望的理解。
7. Introducing Standard Error | 标准误差入门
Even an unbiased estimator will give different estimates from different samples. The standard error of an estimator quantifies how much the estimate is expected to vary from sample to sample. Formally, the standard error (SE) is the standard deviation of the sampling distribution of the estimator. It is a measure of precision: a smaller standard error indicates that the estimator is more reliable.
即使是无偏估计量，不同样本也会给出不同的估计值。估计量的标准误差量化了在不同样本之间该估计值预期波动的程度。正式地说，标准误差（SE）是该估计量的抽样分布的标准差。它是精确度的一种度量：标准误差越小，表明估计量越可靠。
In notation, if we have an estimator θ̂, then SE(θ̂) = √Var(θ̂ to obtain an estimated standard error. For the sample mean, the true standard error is σ/√n, and the estimated standard error is s/√n when σ is unknown.
在记号上，如果我们有一个估计量 θ̂，那么 SE(θ̂) = √Var(θ̂)。标准误差通常依赖于未知的总体参数，因此在实际中我们代入估计值来得到估计的标准误差。对于样本均值，真实的标准误差是 σ/√n，而当 σ 未知时，估计的标准误差是 s/√n。
8. Standard Error of the Mean | 均值的标准误差
d error of the sample mean is perhaps the most frequently encountered standard error. If X̄ is the mean of a random sample of size n from a population with variance σ², then Var(X̄) = σ²/n, and SE(X̄) = σ/√n. This formula is central to constructing confidence intervals and conducting hypothesis tests for a population mean.
样本均值的标准误差可能是最常见的标准误差。如果 X̄ 是来自方差为 σ² 的总体的容量为 n 的随机样本的均值，那么 Var(X̄) = σ²/n，且 SE(X̄) = σ/√n。这个公式是构造置信区间和对总体均值进行假设检验的核心。
stributed with standard deviation 2.5 g. A random sample of 20 packets has a mean weight x̄ = 498 g. Then the standard error of the mean is 2.5/√20 eated the sampling, the sample means would vary with a standard deviation of about 0.559 g around the true population mean.
考虑一个例子：一台机器包装糖包，每袋糖的重量服从正态分布，标准差为2.5克。一个容量为20的随机样本的平均重量为 x̄ = 498克。那么均值的标准误差为 2.5/√20 ≈ 0.559克。这告诉我们，如果重复抽样，各个样本均值将以约0.559克的标准差围绕着真实的总体均值波动。
When σ is unknown, we replace σ with the sample standard deviation s (the square root of the unbiased variance). The estimated standard error is then s/√n. This substitution brings in the t-distribution for small samples, a topic extended in the second year of A-Level.
当 σ 未知时，我们用样本标准差 s（无偏方差的平方根）代替 σ。那么估计的标准误差就是 s/√n。这一替代在小样本下引入了 t 分布，这是A-Level第二年会进一步学习的内容。
9. Standard Error for a Sample Proportion | 样本比例的标准误差
of interest is a population proportion p, the natural estimator is the sample proportion p̂ oulli trials. The sample proportion is an unbiased estimator of p, and its variance is p(1 – p)/n. Therefore, the standard error of p̂ is √[p(1 – p)/n].
当感兴趣的参数是总体比例 p 时，自然的估计量是样本比例 p̂ = X/n，其中 X 是 n 次独立伯努利试验中成功的次数。样本比例是 p 的无偏估计量，其方差为 p(1 – p)/n。因此，p̂ 的标准误差为 √[p(1 – p)/n]。
n most practical situations, the true p is unknown, so the estimated standard error uses p̂: estimated SE = √[p̂(1 – p̂)/n]. For example, in a poll of 200 voters, 110 say they will vote for candidate A. Then p̂ = 110/200 = 0.55. The estimated standard error is √[0.55 margin of uncertainty of about 3.5 percentage points associated with sampling variability.
在大多数实际情况下，真实的 p 是未知的，因此估计的标准误差使用 p̂：估计的 SE = √[p̂(1 – p̂)/n]。例如，在一项对200名选民的调查中，110人表示会投票给候选人A。那么 p̂ = 110/200 = 0.55。估计的标准误差为 √[0.55 × 0.45 / 200] ≈ 0.0352。因此该估计值与抽样变异性相关的不确定性幅度大约为3.5个百分点。
10. Interpreting Standard Error in Context | 结合背景解释标准误差
The standard error is not a measure of the error of a single estimate from the true value; rather, it is a measure of the long-run variability of the estimator. A small standard error means that the estimator yields estimates that are tightly clustered around the parameter, making it precise. However, we must remember that precision is separate from accuracy. An estimator can be precise (low standard error) but biased, so it consistently misses the target.
标准误差并不是度量单个估计值与真实值之间误差的尺子；相反，它是估计量长期变异性的度量。小的标准误差意味着该估计量给出的估计值紧密地聚集在参数周围，使其很精确。但我们必须记住，精确度与准确度是分开的。一个估计量可能很精确（标准误差小）但却是有偏的，从而始终偏离目标。
When writing conclusions in an exam, you should link the standard error to the reliability of the estimate. For instance: “The standard error of 0.56 g indicates that if many samples of size 20 were taken, the sample mean would typically differ from the true mean by about 0.56 g. This level of precision is acceptable for the machine calibration process.”
在考试中写结论时，你应该将标准误差与估计值的可靠性联系起来。例如：“0.56克的标准误差表明，如果抽取许多容量为20的样本，样本均值通常与真实均值相差约0.56克。该精度水平对于机器校准过程是可以接受的。”
11. Bias-Variance Trade-off and Consistency | 偏差-方差权衡与相合性
SE) is lower than that of an unbiased estimator. The MSE combines bias and variance: MSE(θ̂) = Var(θ̂) + [Bias(θ̂ justification for many estimators. The sample mean and the sample variance s² are both consistent.
在更高级的学习中，你会遇到这样的概念：有些有偏估计量的方差可以非常小，以至于其均方误差（MSE）比无偏估计量的还低。MSE 结合了方差和偏差：MSE(θ̂) = Var(θ̂) + [Bias(θ̂)]²。对于A-Level来说，知道相合性（随着 n 增大，估计量收敛到真实参数）是许多估计量的大样本合理性论证就足够了。样本均值和样本方差 s² 都是相合的。
Being able to comment on the behaviour of an estimator as n grows, using the standard error expression, is a useful skill. For example, because the standard error of X̄ is σ/√n n becomes arbitrarily precise and converges to μ in probability, which is the definition of consistency.
能够利用标准误差表达式评论估计量随 n 增大的行为是一项有用的技能。例如，因为 X̄ 的标准误差是 σ/√n，当 n → ∞ 时，SE → 0，这意味着样本均值变得任意精确并以概率收敛到 μ，这就是相合性的定义。
12. Summary and Key Takeaways | 总结与核心要点
Estimators are statistical recipes for guessing population parameters. Bias tells us whether the recipe systematically misses the mark, while standard error tells us how much scatter we can expect from sample to sample. Together, they help us evaluate and choose between estimators. The sample mean and the sample variance using n – 1 are the textbook unbiased estimators, and their standard errors scale with 1/√n, reflecting the power of larger samples.
估计量是猜测总体参数的统计学配方。偏差告诉我们该配方是否系统性地偏离目标，而标准误差告诉我们样本之间预期会有多少波动。两者结合，帮助我们评估并选择估计量。样本均值和使用 n – 1 的样本方差是教科书式的无偏估计量，它们的标准误差以 1/√n 的比例变化，反映了更大样本的威力。
In Edexcel examinations, you should be able to define bias, check unbiasedness using expectation, calculate standard error for means and proportions, and interpret the values in context. Mastery of these fundamentals will make the later topics of confidence intervals and hypothesis testing much more accessible.
在Edexcel考试中，你应该能够定义偏差，使用期望检验无偏性，计算均值和比例的标准误差，并结合背景解释这些数值。掌握这些基础知识将使后续的置信区间和假设检验专题变得更加容易理解。
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