📚 Variance and Standard Deviation: Calculation and Applications | 方差与标准差的计算与应用
When analysing data, the mean alone is not enough. Two datasets can have exactly the same average yet look completely different: one may be tightly clustered around the centre, while the other spreads widely. Variance and standard deviation measure this spread, making them essential tools in A-Level Mathematics, particularly in statistics and probability.
在分析数据时，仅靠平均数远远不够。两组数据可以有完全相同的平均值，但外观却截然不同：一组可能紧密聚集在中心附近，另一组则分布广泛。方差与标准差正是用来衡量这种离散程度的，它们是 A-Level 数学中统计学与概率部分不可或缺的工具。
1. Why Measure Dispersion | 为什么要衡量离散程度
The mean, median and mode describe the central tendency of a dataset, but they say nothing about how spread out the values are. Consider two groups of students: Group A has scores 50, 51, 49, 50, 50; Group B has scores 10, 50, 90, 30, 70. Both have a mean of 50, yet Group B is far more variable. Dispersion measures tell us whether data are consistent or erratic.
平均数、中位数与众数描述了数据集的集中趋势，但无法说明数值的分散情况。设想两组学生：A 组分数为 50、51、49、50、50；B 组为 10、50、90、30、70。两组平均数都为 50，但 B 组的波动性远大于 A 组。离散程度度量告诉我们数据是一致的还是波动的。
Variance and standard deviation are the most common measures of dispersion. They quantify the average distance of each data point from the mean, giving a single number that summarises the entire spread of the data.
方差与标准差是最常见的离散程度度量。它们量化每个数据点与平均数之间的平均距离，用一个数值概括整个数据集的离散状况。
2. Definitions and Key Formulas | 定义与核心公式
For a population with N values x₁, x₂, …, xₙ and mean μ, the population variance σ² is defined as the average of the squared deviations from the mean:
对于包含 N 个数值 x₁, x₂, …, xₙ、平均数为 μ 的总体，总体方差 σ² 定义为各数据点与平均数之差的平方的平均值：
σ² = Σ(x − μ)² ÷ N
The population standard deviation σ is simply the positive square root of the variance:
总体标准差 σ 就是方差的正平方根：
σ = √(Σ(x − μ)² ÷ N)
Why do we square the deviations? Squaring removes negative signs, and it gives greater weight to values that are far from the mean. This makes variance and standard deviation more sensitive to outliers than the mean absolute deviation.
为什么要对偏差进行平方？平方可以消除负号，同时也给予远离平均数的数据更大的权重。这使得方差与标准差比平均绝对偏差对异常值更为敏感。
3. Population vs Sample | 总体与样本的区别
In A-Level statistics, it is crucial to distinguish between a population and a sample. A population includes every member of a group; a sample is a subset chosen to represent the population. When we calculate spread from a sample, we usually want to estimate the population variance, which requires a slightly different formula.
在 A-Level 统计学中，区分总体与样本至关重要。总体包括一个群体中的所有成员；样本则是从中选出的用于代表总体的子集。当我们用样本计算离散程度时，通常希望估计总体方差，这需要使用略有不同的公式。
For a sample of n values with mean x̄, the sample variance is denoted s² and calculated using (n − 1) in the denominator:
对于平均数为 x̄ 的 n 个样本值，样本方差记为 s²，计算时使用 (n − 1) 作为分母：
s² = Σ(x − x̄)² ÷ (n − 1)
The reason for dividing by n − 1 rather than n is called Bessel's correction. Using n tends to underestimate the true population variance, especially for small samples, because the sample mean is itself calculated from the data and is therefore closer to all data points than the true population mean.
使用 n − 1 而非 n 作为除数，称为贝塞尔校正。使用 n 往往会低估真实的总体方差，尤其对于小样本而言，因为样本平均数本身是从数据中计算出来的，因此它比真实的总体平均数更接近所有数据点。
4. Calculating from Raw Data | 原始数据的计算
Given an ungrouped dataset, the most efficient method in an exam is to use the alternative formula:
对于未分组的数据集，考试中最有效的方法是使用替代公式：
σ² = Σx² ÷ N − μ²
value and adding the results; finally substitute into the formula above. The standard deviation is again obtained by taking the square root.
该公式与定义式在数学上等价，但只需要两个和：Σx 与 Σx²。使用步骤如下：首先计算平均数 μ = Σx ÷ N；然后通过将每个值平方并求和得到 Σx²；最后代入上述公式。标准差同样通过取平方根获得。
16 + 49 + 81 + 144 + 324 = 614. Therefore σ² = 614 ÷ 5 − 10² = 122.8 − 100 = 22.8, and σ = √22 omputing each individual deviation separately, saving time and reducing arithmetic errors.
例如，考虑数据集 4、7、9、12、18。N = 5，Σx = 50，因此 μ = 10。同时 Σx² = 16 + 49 + 81 + 144 + 324 = 614。因此 σ² = 614 ÷ 5 − 10² = 122.8 − 100 = 22.8，σ = √22.8 ≈ 4.77。这个方法避免了逐一计算每个偏差，既节省时间又减少算术错误。
5. Frequency Tables | 频率表的处理
When data are presented in a frequency table, each value x appears with a frequency f. The formulas are extended by weighting each value by its frequency:
当数据以频率表的形式呈现时，每个值 x 对应一个频率 f。公式通过将每个值按其频率加权来扩展：
σ² = Σfx² ÷ Σf − (Σfx ÷ Σf)²
llows from the same structural logic as before. This grouping of repeated values makes the calculation much faster than listing every single observation, which is especially beneficial when N is large.
其中 Σf = N 为总频率。平均数 μ = Σfx ÷ Σf，方差遵循与之前相同的结构逻辑。对重复值进行分组使计算远快于逐一列出每个观测值，这在 N 很大时尤为有利。
0 = 304. Then σ² = 304 ÷ 15 − (64 ÷ 15)² ≈ 20.267 − 18.204 = 2.063, giving σ ≈ 1.44 books.
假设一项调查记录学生一个月内阅读的书籍数量：3 名学生读了 2 本，7 名学生读了 4 本，5 名学生读了 6 本。则 Σf = 15，Σfx = 3×2 + 7×4 + 5×6 = 6 + 28 + 30 = 64，所以 μ = 64 ÷ 15 ≈ 4.267。同时 Σfx² = 3×4 + 7×16 + 5×36 = 12 + 112 + 180 = 304。因此 σ² = 304 ÷ 15 − (64 ÷ 15)² ≈ 20.267 − 18.204 = 2.063，σ ≈ 1.44 本。
6. Grouped Data and Midpoints | 分组数据与组中值
Continuous data are often grouped into intervals such as 0–10, 10–20, and so on. In this case, we do not know the exact values within each interval, so we assume every observation lies at the midpoint of its class. The midpoint m is calculated as (upper class boundary + lower class boundary) ÷ 2. The variance formula then becomes:
连续数据通常被分组为区间，如 0–10、10–20 等。在这种情况下，我们不知道每个区间内的精确值，因此假设每个观测值都位于其组段的中点。组中值 m 的计算公式为（上组界 ＋ 下组界）÷ 2。方差公式随后变为：
σ² = Σfm² ÷ Σf − (Σfm ÷ Σf)²
Using midpoints is an approximation, not an exact calculation. If the data within each interval are roughly symmetric around the midpoint, the approximation is good; if the data are heavily skewed within the intervals, the result may be misleading. Always state that the answer is an estimate when working with grouped data.
使用组中值是近似而非精确计算。如果每个区间内的数据大致围绕组中值对称，则近似效果良好；如果区间内数据严重偏斜，结果可能具有误导性。处理分组数据时，务必注明答案为估计值。
7. Interpreting the Standard Deviation | 解读标准差
The standard deviation has the same units as the original data, which makes it easier to interpret than variance. If the standard deviation is small, the data are concentrated close to the mean; if it is large, the data are widely scattered. For approximately symmetric, bell-shaped distributions, the empirical rule states that about 68% of data lie within one standard deviation of the mean, about 95% within two, and about 99.7% within three.
标准差与原始数据具有相同的单位，这使得它比方差更易于解读。如果标准差较小，数据集中在平均数附近；如果较大，则数据分布广泛。对于近似对称的钟形分布，经验法则表明大约 68% 的数据位于平均数的一个标准差范围内，约 95% 位于两个标准差范围内，约 99.7% 位于三个标准差范围内。
Beyond this, Chebyshev's inequality guarantees tthat at least 1 − 1/k² of data lie within k standard deviations of the mean for any distribution, provided k > 1. For example, at least 75% of data always lie within two standard deviations. This is a powerful tool when the shape of the distribution is unknown.
除此之外，切比雪夫不等式保证，对于任何分布，至少有 1 − 1/k² 的数据位于平均数 k 个标准差范围内，只要 k > 1。例如，至少 75% 的数据总是位于两个标准差范围内。当分布形状未知时，这是一个强有力的工具。
8. Applications: Comparing Data Sets | 应用：比较数据集
One of the most common exam questions involves comparing two datasets using the mean and standard deviation. The mean tells us which group performs better on average; the standard deviation tells us which group is more consistent. A lower standard deviation means less variability and therefore greater reliability, while a higher standard deviation indicates more inconsistency and a wider range of outcomes.
最常见的考题之一是使用平均数与标准差比较两组数据。平均数告诉我们哪一组平均表现更好；标准差告诉我们哪一组更稳定。标准差较低意味着变异性较小，因此可靠性更高；标准差较高则表明不一致性更大，结果范围更宽。
For instance, a teacher comparing two classes might find that Class 1 has mean 62 and standard deviation 5, while Class 2 has mean 62 and standard deviation 12. The averages are the same, but Class 1 is far more consistent: most students scored near 62, whereas Class 2 contains both very high and very low scores. This comparison is essential in fields as diverse as finance, medicine and quality control.
例如，比较两个班级的老师可能会发现，1 班平均分为 62、标准差为 5，而 2 班平均分也是 62、标准差为 12。平均数相同，但 1 班要稳定得多：大多数学生得分接近 62，而 2 班既有高分也有低分。这种比较在金融、医学和质量控制等许多领域都至关重要。
9. Linking to Probability Distributions | 与概率分布的联系
of a discrete random variable X with probability distribution P(X = xᵢ) = pᵢ is given by:
方差与标准差在概率中也发挥着核心作用。离散型随机变量 X 的概率分布为 P(X = xᵢ) = pᵢ 时，其方差为：
Var(X) = Σpᵢxᵢ² − (Σpᵢxᵢ)²
√Var(X), and it measures how much the outcomes deviate from the expected value on average.
其中 E(X) = Σpᵢxᵢ 为期望值，E(X²) = Σpᵢxᵢ²。因此 Var(X) = E(X²) − [E(X)]²。这一简洁恒等式非常有用，因为它使我们无需逐一将每个结果减去期望值就能计算方差。随机变量的标准差为 √Var(X)，它衡量结果平均偏离期望值的程度。
10. Common Pitfalls and Exam Tips | 常见错误与应试技巧
Students frequently make several predictable errors. First, forgetting to take the square root: variance and standard deviation are not the same, and an exam question asking for standard deviation will not accept variance. Second, using the wrong denominator: if the data are declared to be a sample, divide by n − 1; if a population, divide by N. Third, confusing the two formulas for variance: remember that Σx² ÷ N − μ² is the computational shortcut, not the definition.
学生经常犯几个可预见的错误。第一，忘记取平方根：方差与标准差并不是一回事，考题要求标准差时不会接受方差答案。第二，用错分母：如果数据注明是样本，则除以 n − 1；如果是总体，则除以 N。第三，混淆方差的两个公式：记住 Σx² ÷ N − μ² 是计算捷径，而非定义式。
Another common issue is using midpoints in grouped data while forgetting that the result is approximate. Also, when calculating Σfx², be careful to square x first and then multiply by f, rather than squaring fx. Finally, always state the units in your final answer, and ensure your final standard deviation is positive: the square root symbol always denotes the positive root.
另一个常见问题是使用分组数据的组中值却忘记结果是近似值。此外，计算 Σfx² 时，务必先对 x 平方再乘以 f，而不是对 fx 整体平方。最后，始终在最终答案中注明单位，并确保最终的标准差为正：平方根符号始终表示正根。
11. Worked Exam Example | 考试例题精讲
+ 3600 + 3844 + 4096 = 18040. Therefore σₐ² = 18040 ÷ 5 − 60² = 3608 − 3600 = 8, and σₐ = √8 ≈ 2.83.
一家公司记录了五天内两台机器的日产量。机器 A 的产量为 56、58、60、62、64 件；机器 B 为 50、55、60、65、70 件。使用平均数与标准差比较两台机器。对于机器 A，Σx = 300，N = 5，因此 μₐ = 60。同时 Σx² = 56² + 58² + 60² + 62² + 64² = 3136 + 3364 + 3600 + 3844 + 4096 = 18040。因此 σₐ² = 18040 ÷ 5 − 60² = 3608 − 3600 = 8，σₐ = √8 ≈ 2.83。
3025 + 3600 + 4225 + 4900 = 18250. Thus σᵦ² = 18250 ÷ 5 − 60² = 3650 − 3600 = 50, and σᵦ = √50 an output of 60 units, but Machine A is far more reliable, with a standard deviation of about 2.83 units compared to Machine B's 7.07 units. Therefore Machine A produces more consistent daily output.
对于机器 B，Σx = 300，μᵦ = 60。同时 Σx² = 50² + 55² + 60² + 65² + 70² = 2500 + 3025 + 3600 + 4225 + 4900 = 18250。因此 σᵦ² = 18250 ÷ 5 − 60² = 3650 − 3600 = 50，σᵦ = √50 ≈ 7.07。两台机器的平均日产量均为 60 件，但机器 A 的可靠性远高于机器 B：其标准差约为 2.83 件，而机器 B 为 7.07 件。因此机器 A 的日产量更为稳定。
12. Summary | 总结
Variance and standard deviation are fundamental measures of dispersion in A-Level Mathematics. The variance is the average of the squared deviations from the mean, while the standard deviation is its square root and shares the same units as the data. The computational formula Σx² ÷ N − μ² simplifies calculations for raw data, and its weighted version applies to frequency tables and grouped data using midpoints. Crucially, population and sample formulas differ in the denominator: N versus n − 1.
方差与标准差是 A-Level 数学中基本的离散程度度量。方差是各数据点与平均数之差的平方的平均值，而标准差则是其平方根，与数据具有相同单位。计算式 Σx² ÷ N − μ² 简化了原始数据的计算，其加权形式适用于频率表和以组中值表示的分组数据。关键在于，总体与样本的公式在分母上有所不同：N 与 n − 1。
These measures are not just mathematical abstractions: they allow us to compare consistency across datasets, to interpret probability distributions, and to make informed decisions in real-world contexts. By mastering the calculation methods, recognising the distinction between population and sample, and avoiding common pitfalls, you will be well prepared for any variance and standard deviation question in your A-Level examination.
这些度量不仅仅是数学抽象：它们使我们能够比较数据集之间的一致性、解释概率分布，并在实际生活中做出明智决策。通过掌握计算方法、认清总体与样本的区别、避免常见错误，你将能从容应对 A-Level 考试中任何关于方差与标准差的问题。
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