📚 Point Estimation (Unbiased Estimators and Estimates) | 点估计（无偏估计量与估计值）
In statistical inference, one of the first tasks we face is using sample data to guess an unknown population parameter. Point estimation provides a single "best guess" value for that parameter. Understanding what makes an estimator good — and particularly what it means for it to be unbiased — is fundamental for IB Mathematics and for any data-driven decision making.
在统计推断中，我们首先面临的任务之一就是利用样本数据去推测未知的总体参数。点估计为那个参数提供一个单一的“最佳猜测”值。理解什么样的估计量是好的——特别是无偏性的真正含义——是 IB 数学和任何基于数据的决策的核心基础。
1. What is Point Estimation? | 什么是点估计？
Point estimation is the process of using a single statistic computed from sample data to estimate an unknown population parameter. For example, if we want to know the average height of all students in a school, we might take a random sample, compute the sample mean, and use that single number as our estimate of the population mean. That sample mean is a point estimator.
点估计是利用样本数据计算出的一个统计量，来估计未知总体参数的过程。例如，如果我们想知道一所学校所有学生的平均身高，我们可以随机抽取一个样本，计算样本均值，并用这个单一的数值作为总体均值的估计。这里的样本均值就是一个点估计量。
In contrast to interval estimation, which gives a range of plausible values, a point estimate is just one number. It might be right on target, or it might be off. The quality of a point estimator is judged by properties such as bias, variance, and consistency.
与给出一个合理范围的区间估计不同，点估计只有一个数值。它可能正中靶心，也可能偏离。点估计量的好坏是通过偏误、方差、一致性等性质来判断的。
2. Population Parameters and Sample Statistics | 总体参数与样本统计量
Before we can discuss estimation, we need to be clear about the notation. A population parameter is a fixed, usually unknown number describing a characteristic of the whole population, such as the population mean μ or population variance σ². A sample statistic, on the other hand, is a function of the observed data and varies from sample to sample. We often place a "hat" on top to denote an estimator, such as μ̂ (estimated μ) or p̂ (estimated proportion).
在讨论估计之前，我们需要清晰地了解符号。总体参数是一个固定的、通常未知的数值，描述整个总体的特征，例如总体均值 μ 或总体方差 σ²。另一方面，样本统计量是观测数据的函数，会随样本的不同而变化。我们通常在字母上方加一个“帽子”来表示估计量，比如 μ̂（μ 的估计值）或 p̂（比例的估计值）。
In IB Mathematics, you will encounter notation like X̄ for the sample mean and s² for the sample variance. These are sample statistics that serve as point estimators for μ and σ² respectively.
在 IB 数学中，你会遇到诸如 X̄ 表示样本均值，s² 表示样本方差等符号。这些样本统计量分别作为总体均值 μ 和总体方差 σ² 的点估计量。
3. Estimators and Estimates | 估计量与估计值
an estimator and an estimate. An estimator is a rule or formula that tells you how to calculate the quantity from sample data. For example, the formula X̄ lug in the actual observed numbers from a particular sample, the resulting number is called an estimate. So the estimator is a random variable (before sampling), while the estimate is a constant (after sampling).
区分估计量与估计值至关重要。估计量是一个规则或公式，告诉你如何从样本数据中计算数量。例如，公式 X̄ = (1/n)∑Xi 就是一个估计量。一旦你把特定样本中实际观测到的数值代入，所得的数字就称为估计值。因此，估计量是一个随机变量（抽样之前），而估计值是一个常数（抽样之后）。
This distinction matters when we talk about bias, because the expected value is taken over the sampling distribution of the estimator, not for a single realized estimate.
当我们讨论偏误时，这一区别很重要，因为期望是针对估计量的抽样分布而言的，而不是针对某一个已得的具体估计值。
4. The Concept of Unbiasedness | 无偏性的概念
An estimator θ̂ is said to be an unbiased estimator of a parameter θ if the expected value of θ̂ equals θ for all possible values of θ. In symbols: E(θ̂) = θ. If E(θ̂) ≠ θ, then θ̂ is biased, and the bias is defined as Bias(θ̂) = E(θ̂) - θ.
若估计量 θ̂ 对所有可能的 θ 值，都满足 E(θ̂) = θ，则称 θ̂ 是参数 θ 的一个无偏估计量。如果用符号表示就是：E(θ̂) = θ。如果 E(θ̂) ≠ θ，那么 θ̂ 就是有偏的，偏误的定义为 Bias(θ̂) = E(θ̂) – θ。
Unbiasedness means that, if we were to repeat the sampling process many times and compute the estimate each time, the average of all those estimates would equal the true parameter. It does not guarantee that a single estimate is correct, but it assures us the estimator is centered on the target in the long run.
无偏性意味着，如果我们多次重复抽样过程并每次都计算估计值，所有那些估计值的平均值会等于真实的参数。这并不能保证单次估计就是正确的，但它确保我们在长期中，估计量以目标为中心。
For IB exams, you should be able to identify whether a given estimator is unbiased, and prove unbiasedness for standard estimators like the sample mean and the sample variance (with n-1 denominator).
对于 IB 考试，你应该能够识别一个给定的估计量是否无偏，并会证明标准估计量如同样本均值和样本方差（分母 n-1）的无偏性。
5. Unbiased Estimator of the Population Mean | 总体均值的无偏估计量
The most commonly used estimator for the population mean μ is the sample mean X̄ = (1/n)∑Xi observations are independent and identically distributed (i.i.d.) with mean μ, we have E(X̄) = μ. Therefore, X̄ is an unbiased estimator of μ.
总体均值 μ 最常使用的估计量是样本均值 X̄ = (1/n)∑Xi。假设观测值是独立同分布的，且具有均值 μ，那么 E(X̄) = μ。因此，X̄ 是 μ 的无偏估计量。
bution with finite mean. That is a direct consequence of the linearity of expectation: E(X̄) = (1/n)∑E(Xi) = (1/n) · nμ = μ.
这个性质不依赖于总体分布是正态的；对于任何有限均值的分布都成立。这是期望线性性质的直接结果：E(X̄) = (1/n)∑E(Xi) = (1/n)·nμ = μ。
Thus, the sample mean is always an unbiased estimator of the population mean. This is one reason why we use X̄ so frequently.
因此，样本均值总是总体均值的无偏估计量。这就是我们频繁使用 X̄ 的原因之一。
6. Proving the Sample Mean is Unbiased | 证明样本均值是无偏的
Let X₁, X₂, …, Xₙ be a random sample from a population with mean μ. Define X̄ = (1/n)∑Xi. Then
E(X̄) = E[(1/n)∑Xi] = (1/n)∑E(Xi) = (1/n)·nμ = μ.
令 X₁, X₂, …, Xₙ 为来自均值为 μ 的总体的一组随机样本。定义 X̄ = (1/n)∑Xi。那么
E(X̄) = E[(1/n)∑Xi] = (1/n)∑E(Xi) = (1/n)·nμ = μ.
Hence the sample mean is unbiased for μ. No assumption about the population variance or distribution shape is needed.
因此，样本均值是 μ 的无偏估计量，无需关于总体方差或分布形状的任何假设。
However, the estimator X̄ itself has a variance: Var(X̄ ill have large variability. We often need to consider both bias and variability when choosing an estimator.
然而，估计量 X̄ 本身具有方差：Var(X̄) = σ²/n。无偏估计量仍然可以有较大的变异性。在选择估计量时，我们通常需要同时考虑偏误和变异性。
7. Unbiased Estimator of the Population Variance | 总体方差的无偏估计量
ation variance σ², a natural guess might be to use the sample variance defined as (1/n)∑(Xi – X̄)². However, this estimator is biased. To obtain an unbiased estimator, we must divide by n–1 instead of n. That gives the unbiased sample variance:
s² = [1/(n–1)] ∑(Xi – X̄)².
对于总体方差 σ²，一个自然的猜测可能是使用定义为 (1/n)∑(Xi – X̄)² 的样本方差。但这个估计量是有偏的。要得到无偏估计量，我们必须除以 n–1 而不是 n。这就得到了无偏样本方差：
s² = [1/(n–1)] ∑(Xi – X̄)².
We then have E(s²) = σ². The factor n– f the estimate, and it accounts for the fact that we used the data to estimate μ through X̄, reducing the effective number of independent pieces of information.
于是我们有 E(s²) = σ²。因子 n–1 被称为估计的自由度，它体现了我们使用了数据通过 X̄ 估计了 μ，从而减少了独立信息的有效数目这一事实。
It is a common IB exam question to explain why we use n–1. The short answer: because it makes the estimator unbiased.
解释为什么使用 n–1 是 IB 考试中的常见问题。简短的回答是：因为这样会让估计量无偏。
8. Why Divide by n-1? A Simple Proof | 为什么除以 n-1？简单证明
To prove E(s²) = σ², start with the sum of squared deviations: S = ∑(Xi – X̄)² = ∑Xi² – nX̄². Then take expectations:
E(S) = ∑E(Xi²) – nE(X̄²).
要证明 E(s²) = σ²，可从平方偏差和入手：S = ∑(Xi – X̄)² = ∑Xi² – nX̄²。然后求期望：
E(S) = ∑E(Xi²) – nE(X̄²).
Using the identity Var(Y) = E(Y²) – [E(Y)]², we get E(Xi²) = σ² + μ² and E(X̄²) = Var(X̄) + μ² = σ²/n + μ². Substitute:
利用等式 Var(Y) = E(Y²) – [E(Y)]²，可得 E(Xi²) = σ² + μ²，并且 E(X̄²) = Var(X̄) + μ² = σ²/n + μ²。代入得：
E(S) = n(σ² + μ²) – n(σ²/n + μ²) = nσ² + nμ² – σ² – nμ² = (n–1)σ².
Therefore, E[S] = (n–1)σ², which implies E[S/(n–1)] = σ². Thus s² is unbiased.
因此，E[S] = (n–1)σ²，这意味着 E[S/(n–1)] = σ²。所以 s² 是无偏的。
If we had used the divisor n, the expectation would be (n–1)/n σ², which is always slightly smaller than σ² — a negative bias.
如果我们使用除数 n，期望值将是 (n–1)/n σ²，它始终略小于 σ²——一种负偏差。
9. Other Examples of Unbiased Estimators | 其他无偏估计量的例子
Beyond the mean and variance, there are many other parameters for which unbiased estimators exist. Some common ones include:
- For a population proportion p, the sample proportion p̂ = X/n (where X is the number of successes) is unbiased: E(p̂) = p.
- For the difference between two population means μ₁ – μ₂, the statistic X̄₁ – X̄₂ is unbiased because E(X̄₁ – X̄₂) = μ₁ – μ₂.
- For a regression slope β₁ in simple linear regression, the least squares estimator β̂₁ is unbiased under the standard assumptions.
除了均值与方差外，还有许多参数存在无偏估计量。一些常见的例子包括：
- 对于总体比例 p，样本比例 p̂ = X/n（X 为成功次数）是无偏的：E(p̂) = p。
- 对于两个总体均值之差 μ₁ – μ₂，统计量 X̄₁ – X̄₂ 是无偏的，因为 E(X̄₁ – X̄₂) = μ₁ – μ₂。
- 在简单线性回归中，回归斜率 β₁ 的最小二乘估计量 β̂₁ 在标准假设下是无偏的。
However, not every parameter has an unbiased estimator that behaves well. For instance, the sample standard deviation s is not an unbiased estimator of σ; it tends to underestimate σ slightly for small samples, although the bias diminishes as n grows.
然而，并非每个参数都有表现良好的无偏估计量。例如，样本标准差 s 不是 σ 的无偏估计量；对于小样本，它往往略微低估 σ，尽管随着 n 增大偏误会消失。
10. Limitations of Unbiasedness and Final Tips | 无偏性的局限与最后提示
ightly biased estimator with much smaller variance can have a lower MSE than an unbiased estimator with large variance. This trade-off is the reason why in some contexts (like regularization in machine learning) we deliberately introduce bias to reduce variance.
无偏性只是估计量众多优良性质中的一个。统计决策理论的一个关键原则是，我们通常关注整体准确度，用均方误差（MSE）衡量。MSE = 方差 + 偏误²。一个稍微有偏但方差小得多的估计量，可能比一个无偏但方差很大的估计量具有更低的 MSE。这种权衡就是在某些情境下（如机器学习中的正则化）我们故意引入偏误以减少方差的原因。
In IB exams, you may be asked to compare two estimators by computing their variances and biases. Always check whether the estimator is centered on the parameter. Also remember: unbiasedness is a property of the estimator's sampling distribution, not a guarantee for a single estimate. Never claim that "the estimate is unbiased," because once you have a number, it either equals the parameter or not. It is the estimator that is unbiased.
在 IB 考试中，你可能会被要求通过计算方差和偏误来比较两个估计量。始终检查估计量是否以参数为中心。还有，请记住：无偏性是估计量抽样分布的性质，而不保证单次估计。永远不要声称“这个估计值是无偏的”，因为一旦你得到了一个数字，它要么等于参数要么不相等。无偏的是估计量。
A final tip: when proving unbiasedness, write the expectation step-by-step, use linearity of expectation, and be careful with the substitution of variance identities. Show clearly the difference between n and n-1 cases.
最后一个提示：证明无偏性时，要逐步写出期望，使用期望的线性性质，并小心代入方差的等式。清晰展示 n 与 n-1 情况之间的区别。
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