Outliers | 异常值 - aleveler.com

📚 Outliers | 异常值

In any data set, we sometimes encounter values that seem far away from the rest of the data. These unusual points are called outliers. They can occur by chance, due to measurement error, or because the sample contains a genuinely extreme observation. Identifying and understanding outliers is a key skill in A‑Level Edexcel Mathematics, as they can heavily influence statistical summaries and conclusions. This article explains what outliers are, how to detect them using the IQR and standard deviation methods, and how to handle them in box plots, measures of location, and bivariate data.

在任何数据集中，我们有时会遇到远离其他数据的数值。这些异常的点被称为异常值。它们可能偶然出现，可能源自测量误差，也可能因为样本本身就包含一个真正的极端观测值。在 A‑Level Edexcel 数学中，识别与理解异常值是一项关键技能，因为它们会严重影响统计汇总与结论。本文将解释什么是异常值，如何使用四分位距法和标准差法检测它们，以及如何在箱线图、位置度量和双变量数据中处理异常值。

1. What Are Outliers? | 什么是异常值？

An outlier is a data point that lies an abnormal distance from other values in a random sample from a population. There is no universal definition, but most textbooks describe an outlier as a value that does not follow the general pattern of the data. In a rough sense, an outlier is a value that is either much smaller than the lower quartile or much larger than the upper quartile by a considerable margin.

异常值是指在一个来自总体的随机样本中，与其他数值存在异常距离的数据点。没有一个通用的定义，但大多数教材将异常值描述为不遵循数据总体模式的值。粗略地说，异常值就是比下四分位数小很多或比上四分位数大很多的数值。

The presence of outliers can distort the mean, inflate the variance, and change the shape of a distribution. However, not every extreme value is a mistake; some outliers carry important information about long tails or rare events. Therefore, in Edexcel exams, you must be able to identify outliers, describe their effect on statistics, and decide whether to keep or remove them based on context.

异常值的存在会歪曲均值、放大方差并改变分布形状。然而，并非所有极端值都是错误；有些异常值包含关于长尾或稀有事件的重要信息。因此，在 Edexcel 考试中，你必须能够识别异常值，描述它们对统计量的影响，并根据上下文决定是保留还是剔除它们。

2. The IQR Rule for Outliers | 四分位距法则检测异常值

The most common method to identify outliers at A‑Level is the interquartile range (IQR) rule. First, find the lower quartile Q₁ and the upper quartile Q₃. The IQR is Q₃ – Q₁. Then calculate the lower fence and upper fence: lower fence = Q₁ – 1.5 × IQR, upper fence = Q₃ + 1.5 × IQR. Any data value less than the lower fence or greater than the upper fence is considered an outlier.

A‑Level 中最常用的识别异常值方法是四分位距 (IQR) 法则。首先找到下四分位数 Q₁ 和上四分位数 Q₃。IQR = Q₃ – Q₁。然后计算下限值和上限值：下限 = Q₁ – 1.5 × IQR，上限 = Q₃ + 1.5 × IQR。任何小于下限或大于上限的数据值都被视为异常值。

Sometimes a second, stricter boundary is used: the lower outer fence Q₁ – 3 × IQR and upper outer fence Q₃ + 3 × IQR. Values beyond these are called extreme outliers. In Edexcel papers, you usually only need to apply the 1.5 × IQR rule, but mentioning the 3 × IQR rule can show deeper understanding.

有时还会使用第二条更严格的边界：下外限 Q₁ – 3 × IQR 和上外限 Q₃ + 3 × IQR。超出这些边界的值被称为极端异常值。在 Edexcel 试卷中，通常只需应用 1.5 × IQR 规则，但提及 3 × IQR 规则可以展示更深的理解。

3. Worked Example: IQR Method | 实例：四分位距法

Consider the data set: 2, 5, 7, 9, 10, 12, 14, 30. Find the quartiles: for n = 8, Q₁ is at position 8/4 = 2, so Q₁ = 5; Q₃ is at position 3×8/4 = 6, so Q₃ = 12. (Edexcel may use different interpolation rules; here we use the simple position method for discrete data.) Then IQR = 12 – 5 = 7. Lower fence = 5 – 1.5 × 7 = 5 – 10.5 = –5.5. Upper fence = 12 + 1.5 × 7 = 12 + 10.5 = 22.5. The value 30 is above 22.5, so it is an outlier.

考虑数据集：2, 5, 7, 9, 10, 12, 14, 30。找到四分位数：n = 8，Q₁ 位置为 8/4 = 2，即 Q₁ = 5；Q₃ 位置为 3×8/4 = 6，即 Q₃ = 12。（Edexcel 可能使用不同的插值规则；此处我们对离散数据使用简单位置法。）则 IQR = 12 – 5 = 7。下限 = 5 – 1.5 × 7 = 5 – 10.5 = –5.5。上限 = 12 + 1.5 × 7 = 12 + 10.5 = 22.5。数值 30 大于 22.5，因此它是异常值。

Notice that the lower fence is negative, but the data are all positive, so no low‑end outlier exists. In an exam, always clearly state the fences, show the comparison, and write a conclusion.

请注意下限为负数，但所有数据均为正值，因此不存在低端异常值。在考试中，一定要明确写出上下限，展示比较过程，并写出结论。

4. Outliers and Box Plots | 异常值与箱线图

Box plots (box‑and‑whisker diagrams) are closely linked to outliers. A standard box plot shows the minimum, Q₁, median, Q₃, and maximum. When outliers are present, the whiskers are often drawn only up to the most extreme value that is not an outlier, and outliers are plotted as individual points (usually crosses or dots) beyond the whiskers.

箱线图与异常值密切相关。标准箱线图显示最小值、Q₁、中位数、Q₃ 和最大值。当存在异常值时，须线通常只延伸到非异常值中的最极端值，而异常值则绘制为须线之外的单独点（通常用叉号或圆点表示）。

When asked to draw a box plot with outliers, first identify the fences. Then locate the highest value inside the upper fence (the upper adjacent value) and the lowest value inside the lower fence (the lower adjacent value). These become the endpoints of the whiskers. Mark any data beyond the fences as outliers. This modified box plot gives a clearer picture of the central 50% of the data, free from the influence of extreme values.

当要求绘制带有异常值的箱线图时，首先确定上下限。然后找出上限内的最大值（上相邻值）和下限内的最小值（下相邻值）。这些值成为须线的端点。将任何超出上下限的数据标记为异常值。这种修正箱线图能更清晰地展示不受极端值影响的中间 50% 数据。

5. Standard Deviation and Outliers | 标准差与异常值

Another way to spot outliers is to use the mean and standard deviation. For a roughly symmetric distribution, a common rule is that data values more than 2 standard deviations from the mean are possible outliers, and values more than 3 standard deviations from the mean are extreme outliers. In Edexcel, you might be asked to consider whether a value more than 2 s.d. or 3 s.d. from the mean is an outlier.

另一种发现异常值的方法是使用均值与标准差。对于大致对称的分布，一个常用规则是，距离均值超过 2 个标准差的数值可能是异常值，距离均值超过 3 个标准差的数值是极端异常值。在 Edexcel 考试中，你可能会被问及距离均值超过 2 个或 3 个标准差的数值是否属于异常值。

For example, if the mean is 50 and the standard deviation is 8, the interval μ ± 2σ is 50 ± 16, i.e. 34 to 66. A value of 70 lies outside this range and could be flagged as an outlier. This method is less robust than the IQR rule because the mean and standard deviation themselves are affected by outliers. Therefore, the IQR method is preferred when the data may contain extreme values.

例如，若均值为 50，标准差为 8，则区间 μ ± 2σ 为 50 ± 16，即 34 到 66。数值 70 在此范围之外，可被标记为异常值。此方法不如四分位距法稳健，因为均值和标准差本身会受到异常值的影响。因此，当数据可能包含极端值时，IQR 方法更为可取。

6. Impact on Mean and Standard Deviation | 对均值和标准差的影响

Outliers can dramatically change the sample mean. Because the mean uses every value, a single extremely large value pulls the mean upward, while an extremely small value drags it downward. The median, however, is resistant to outliers because it depends only on the middle position. This is why the median is often called a robust statistic.

异常值可以极大地改变样本均值。由于均值使用每一个数值，一个极大的值会将均值向上拉，而一个极小的值则会将它向下拖。然而，中位数对异常值有抵抗力，因为它只依赖于中间位置。这就是为什么中位数常被称为稳健统计量。

The standard deviation is also inflated by outliers because it squares deviations from the mean. A far‑out value creates a large squared deviation, increasing the variance and standard deviation considerably. When outliers are removed, both the mean and standard deviation shift closer to the bulk of the data, often giving a better summary of the typical behaviour.

标准差也会因异常值而膨胀，因为它将偏离均值的差平方。一个远离的值会产生很大的平方差，从而显著增大方差和标准差。当移除异常值后，均值和标准差都会向数据主体靠拢，通常能更好地概括典型行为。

7. Outliers in Bivariate Data | 双变量数据中的异常值

In scatter diagrams and correlation/regression analysis, an outlier is a point that does not follow the general trend of the rest of the data. It might have an unusual combination of x and y values. Even if neither the x nor y value alone is extreme, the pair can be an outlier in the context of the relationship. In Edexcel, you may be asked to describe the effect of such a point on the product moment correlation coefficient (PMCC) or the regression line.

在散点图和相关/回归分析中，异常值是指不遵循其余数据总体趋势的点。它可能具有不寻常的 x 与 y 组合。即使 x 或 y 值单独看并不极端，但在关系背景下，该组合点仍可成为异常值。在 Edexcel 考试中，你可能会被要求描述这样的点对积矩相关系数 (PMCC) 或回归线的影响。

An outlier can weaken or strengthen the apparent correlation, depending on its position. A point far from the line of best fit tends to reduce the absolute value of r. On the other hand, a point that lies along the extension of the trend but far from the other points can artificially inflate the correlation coefficient. When constructing a regression line, one outlier can pull the line towards itself, altering the slope and intercept substantially.

异常值根据其位置可以削弱或增强表观相关性。远离最佳拟合线的点往往会降低 |r| 的值。另一方面，若一个点位于趋势延长线上但远离其他点，则可能人为地夸大相关系数。在构建回归线时，一个异常值能将线条拉向自身，显著改变斜率和截距。

8. Handling Outliers: To Remove or Not? | 异常值处理：删除与否？

The decision to remove an outlier should never be automatic. In Edexcel problems, you must use the context. If an outlier is due to a measurement error or data entry mistake, it can be corrected or removed. If it is a genuine extreme but rare occurrence, you might keep it and note its influence. Sometimes it is useful to perform the analysis both with and without the outlier to report the difference.

移除异常值的决定绝不应是自动的。在 Edexcel 题目中，你必须根据上下文判断。如果异常值源自测量错误或数据录入失误，可以对其进行修正或删除。如果它是一次真实的极端但罕见的事件，你可能要保留它并说明其影响。有时，分别在使用和不使用异常值的情况下进行分析并报告差异会很有用。

Removing an outlier without justification can lead to loss of information. For example, in a clinical trial, an extreme reaction might be the most important observation. Always give a reason: “The value 30 is an outlier based on the IQR rule, and since it was recorded incorrectly, it will be removed from further calculations.”

无理由地删除异常值可能导致信息丢失。例如，在临床试验中，一次极端反应可能是最重要的观察结果。一定要给出理由：“根据 IQR 法则，数值 30 是一个异常值，并且由于它记录有误，将在后续计算中被移除。”

9. Exam Tips for Edexcel | Edexcel 考试技巧

Show all steps when identifying outliers: state Q₁, Q₃, IQR, then the fences, and finally list the outliers. Clearly label outliers on a box plot using crosses or small circles. The Edexcel mark scheme often awards marks for the correct fences and for plotting outliers explicitly.

识别异常值时展示所有步骤：写出 Q₁、Q₃、IQR，然后写出上下限，最后列出异常值。在箱线图上使用叉号或小圆圈清晰地标出异常值。Edexcel 评分方案通常对正确的上下限和明确绘制异常值给予分数。

When explaining the effect on the mean or correlation, be precise: say “the outlier increases the mean because it is a large value” rather than just “it changes the mean.” Use the words “resistant” for median and IQR, and “sensitive” for mean and standard deviation. If asked to clean data, always justify your choice with a statistical rule and context.

在解释对均值或相关性的影响时，要精确：说“因为异常值是一个大数值，所以它增大了均值”，而不仅仅是“它改变了均值”。对中位数和 IQR 使用“稳健”一词，对均值和标准差使用“敏感”一词。如果被要求清理数据，一定要用统计规则和上下文为你的选择提供理由。

10. Common Mistakes to Avoid | 常见错误避免

One frequent mistake is forgetting to multiply the IQR by 1.5 when calculating fences. Another is misidentifying quartile positions: for small data sets, ensure you use the correct method (Edexcel often expects the median to be included or excluded according to the convention taught). Also, students sometimes report the fences as outliers themselves — remember, outliers are the data values that fall outside the fences.

一个常见错误是在计算上下限时忘记将 IQR 乘以 1.5。另一个是错误定位四分位数：对于小数据集，确保使用正确的方法（Edexcel 通常期望按照所教惯例包含或排除中位数）。此外，学生有时会报告上下限本身为异常值——请记住，异常值是落在上下限之外的数据值。

Be careful not to confuse the outlier boundary 1.5 × IQR with 1.5 times the standard deviation. Also, when drawing a box plot with outliers, do not extend the whiskers to the fences — extend them to the maximum data value inside the fences. Finally, do not remove outliers simply because they “look weird”; always back up your decision with calculation and reasoning.

注意不要将异常值边界 1.5 × IQR 与 1.5 倍标准差混淆。此外，在绘制带异常值的箱线图时，不要将须线延伸到上下限——应延伸到上下限内的最大数据值。最后，不要仅仅因为异常值“看起来奇怪”就删除它们；始终用计算和推理支持你的决定。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）

Outliers | 异常值