Historical Context & Motivation
Humans have been summarizing data for centuries. Ancient civilizations recorded harvests and trade goods, but the idea of calculating a single number to represent an entire collection of values is surprisingly modern. The concept of the arithmetic mean (what most people call the "average") became a formal mathematical tool during the Scientific Revolution, while the median gained recognition later as statisticians realized that extreme values could warp averages in misleading ways.
The central question this lesson addresses is straightforward yet powerful: when your data set contains one or more extreme values—called outliers—which measure of center, the mean or the median, gives a more accurate picture of the 'typical' value? Understanding this distinction is essential for interpreting reports about income, test scores, housing prices, and nearly any real-world data set.
Core Principles & Definitions
Before we can explore how outliers pull or push summary statistics, we need clear definitions of the key terms involved. Each concept below builds on ideas you already know from Algebra 1 and basic data handling.
Mean
Median
Outlier
Resistant Measure
Skewness
Visual Explanation — Dot Plots With and Without Outliers
The diagram below shows two dot plots using the same core data set: {2, 3, 4, 5, 6}. In the top plot, no outlier is present and the mean and median coincide at 4. In the bottom plot, the value 6 has been replaced with 20—an outlier. Notice how the mean jumps from 4 to 6.8 while the median stays at 4.
The visual makes a critical idea concrete: because the mean uses every value in its calculation, any extremely large or small number has a direct arithmetic impact. The median ignores how far away extreme values are—it only cares about which value sits in the middle position once the data are sorted.
Mathematical Framework
Let's formalize how the mean and median are calculated so you can see exactly why one is sensitive to outliers and the other is not.
The key insight is algebraic: in the mean formula, changing any single value directly changes the numerator by the same amount, which then propagates through the division. In the median formula, changing an extreme value has zero effect on the sorted position of the middle value. This is why statisticians call the median a resistant measure of center.
Outliers, Skewness, and Choosing the Right Measure
Outliers don't just move the mean—they can change the overall shape of a distribution. When a data set has one or more extreme high values, we say the distribution is skewed right (the tail stretches to the right). With extreme low values, it is skewed left. When data are roughly symmetric, the mean and median are close together and either works well. The diagram below contrasts three distribution shapes.
| Distribution Shape | Mean vs. Median | Better Measure of Center |
|---|---|---|
| Symmetric (no outliers) | Mean ≈ Median | Either (mean is common) |
| Right-skewed (high outliers) | Mean > Median | Median |
| Left-skewed (low outliers) | Mean < Median | Median |
Worked Example — Test Scores
A class of nine students takes a quiz. The scores are: 72, 75, 78, 80, 82, 84, 85, 88, 150. The score of 150 looks suspiciously high (perhaps the teacher entered the wrong grade). Let's compute the mean and median, then decide which one better represents the typical student score.
Strengths & Limitations — Mean vs. Median
| Feature | Mean | Median |
|---|---|---|
| Sensitivity to outliers | High — one extreme value changes it significantly | Low — remains stable even with extreme values (resistant) |
| Uses all data values | Yes — every value contributes to the sum | No — only the middle value(s) matter |
| Best for symmetric data | Excellent — equals the median and is easy to compute | Also good, but provides the same answer as the mean |
| Best for skewed data | Misleading — pulled toward the tail | Preferred — stays near the bulk of the data |
| Common real-world use | GPA, batting average, temperature averages | Household income, home prices, hospital wait times |
Connection to Advanced Statistics
The ideas you've learned here extend into more advanced courses in surprising ways. Understanding how outliers affect the center prepares you for topics like robust regression, trimmed means, and non-parametric statistics—all of which deal with data that don't follow neat, symmetric patterns.
| Concept in This Lesson | Advanced Extension |
|---|---|
| Mean is sensitive to outliers | Trimmed mean (drop top/bottom 5–10% of data) combines resistance with the mean's mathematical usefulness |
| Median is resistant | Median Absolute Deviation (MAD) extends resistance to measures of spread |
| IQR rule for outliers | Z-score method (|z| > 2 or 3) is used in AP Statistics and beyond for more formal outlier detection |
| Choosing mean vs. median | In inference, non-parametric tests like the Wilcoxon rank-sum test use medians instead of means when assumptions are violated |
For now, the most important skill is being able to justify your choice of measure. On tests and projects, stating 'the median is better because outliers pull the mean away from the center of the data' earns full credit. In more advanced courses, you'll learn formal ways to quantify how much influence an outlier has.
Practice Problems
Lesson Summary
An outlier is a data value that is far removed from the rest of the data. Because the mean uses every value in its calculation (sum ÷ count), a single extreme number can pull the mean significantly toward itself. The median, defined as the middle value of sorted data, depends only on position and is therefore a resistant measure of center that remains stable despite outliers.
When a distribution is skewed by outliers, the median is usually the better measure of center because it more accurately represents the typical value. In symmetric distributions without outliers, the mean and median are approximately equal and either may be used. To justify your choice on an assessment, state whether outliers are present, describe how they affect the mean, and explain why the median is or is not a better representation of the data's center.