Let the Distribution Guide Your Choice
In Describing Center in Context, you learned to report a mean or median for a named group and variable, with units. The next question is which measure better represents the center of a particular distribution. The shape of the distribution and the presence of unusual values help you decide.
A distribution that is approximately symmetric and has no strong outliers is often well summarized by its mean. When a distribution is skewed or contains an extreme value, the median is often more representative of a typical observation. These are useful guidelines, not rules that make the other measure incorrect in every situation. Your choice should fit the shape, the data, and the question being asked.
The mean is the arithmetic average, and the median is the middle value in the ordered data (or the average of the two middle values when there is an even number of observations). You do not need to calculate both every time. Looking at a dotplot, histogram, or stemplot first can show whether one measure is likely to give a more useful summary.
A Practical Decision Process
As in The SOCS Framework for Describing Distributions, examine shape and outliers before settling on center. A quick decision process is:
Use an appropriate graph to decide whether the pattern is roughly symmetric, skewed, or affected by an observation that stands apart.
If the distribution is approximately symmetric and has no strong outliers, the mean can describe its balance point and uses every observation.
If the distribution is skewed or has an extreme value, the median often better describes the middle of the observations because a few extreme values have limited influence on it.
Name the measure, report its value and units, and connect your choice to the distribution’s shape or unusual values.
“Resistant” does not mean completely unchanged under every alteration to a data set. It means that the median is less affected by extreme values than the mean generally is. An extreme value can move the mean substantially because it contributes fully to the total used to calculate the average. The median depends on the ordered position of the observations, so one very large value may not change the middle position at all.
Shape matters even when no value is formally flagged as an outlier. In a skewed distribution, observations extend farther on one side than the other. The mean is pulled toward that longer tail, so the median may better mark the center of the main body of observations. In a roughly symmetric distribution, there is not a long tail pulling the mean away from the middle, and the mean and median are often fairly close.
Worked Examples
Worked Example: Center of Balanced Daily Temperatures
A fictional weather station records these eight daily high temperatures, in degrees Celsius: \(16, 18, 19, 20, 20, 21, 22,\) and \(24\). The values are balanced around 20, with no observation standing apart. Which measure of center is a sensible choice, and what does it say?
State. The distribution is approximately symmetric and has no strong outlier. The mean is a sensible measure of center.
Plan. Find the mean by adding the eight temperatures and dividing by 8. Find the median from the two middle ordered observations. Compare the results and interpret the selected center in context.
Do. The temperatures are already in order. Their sum is \(16+18+19+20+20+21+22+24=160\) degrees Celsius, so the mean is:
With eight observations, the median is the average of the fourth and fifth values. Both are \(20^\circ\text{C}\), so the median is \((20+20)/2=20^\circ\text{C}\). As a check on the sum, the values can be paired around 20: \(16+24=40\), \(18+22=40\), \(19+21=40\), and \(20+20=40\). The total is \(160\).
Conclude. The mean daily high temperature in these eight fictional records is \(20^\circ\text{C}\). Because the distribution is balanced and has no strong outlier, the mean provides a reasonable description of its center. Here the mean and median agree, which is consistent with the balanced pattern.
Worked Example: One Extreme Appointment Wait
A fictional clinic records these nine appointment waits, in minutes: \(18, 19, 20, 20, 21, 22, 23, 24,\) and \(120\). The first eight waits are relatively close together, while one wait is much longer. Compare the mean and median, then choose a useful center for a typical wait.
State. The distribution has a very large value far from the others, creating a long right tail. The median is likely to better represent a typical wait than the mean.
Plan. Calculate the mean using all nine waits and identify the fifth value as the median. To see how much the extreme observation matters, also compare the mean and median for the first eight waits without it. The nine recorded values are the data to summarize; the comparison shows how the extreme value affects the two measures.
Do. The first eight waits total \(167\) minutes. Including the \(120\)-minute wait gives a total of \(287\) minutes, so:
In the ordered list of nine waits, the fifth value is \(21\), so the median is \(21\) minutes. For the first eight values, the mean is \(167/8=20.875\), or about \(20.9\) minutes, and the median is the average of the fourth and fifth values: \((20+21)/2=20.5\) minutes. Thus, adding the \(120\)-minute wait raises the mean from about \(20.9\) to \(31.9\) minutes, while the median moves only from \(20.5\) to \(21\) minutes. The total can be checked as \(18+19+20+20+21+22+23+24=167\), then \(167+120=287\).
Conclude. The median wait among these nine fictional appointments is \(21\) minutes. Because one unusually long wait pulls the mean upward, the median is a more useful summary of a typical wait in this distribution. The mean of about \(31.9\) minutes is still mathematically correct; it is simply less representative of the cluster of waits near 18 to 24 minutes.
Worked Example: Skew Without a Stand-Alone Outlier
A fictional sample of nine technical support tickets took \(4, 5, 6, 7, 8, 9, 10, 12,\) and \(14\) minutes to resolve. The values become less frequent toward the higher end, making the distribution right-skewed, but no single observation stands far apart from the rest. Which center would you report for a typical resolution time?
State. The distribution is right-skewed, with a gradual extension toward higher times rather than one isolated extreme value. The median is a reasonable choice for describing a typical resolution time.
Plan. Find the mean by dividing the sum of the nine times by 9. Since the ordered list has nine values, its median is the fifth value. Compare the two values, then use the shape to explain the choice.
Do. The sum of the times is \(4+5+6+7+8+9+10+12+14=75\) minutes. Therefore:
The fifth value in the ordered list is \(8\), so the median is \(8\) minutes. A check on the total is \(4+5+6+7=22\), \(8+9+10=27\), and \(12+14=26\); \(22+27+26=75\).
Conclude. The median resolution time in this fictional sample is \(8\) minutes. The distribution’s right tail pulls the mean slightly above the median, so the median gives a clear description of the middle time without being pulled toward the higher values. The mean, about \(8.33\) minutes, is also a valid summary, but the skew makes the median a useful choice for a typical ticket.
Connecting Center to Spread
A center measure is more informative when you consider the distribution’s variability, too. As you learned in Describing Spread in Context, different measures of spread summarize different features of the data. A common pairing is the mean with standard deviation for a roughly symmetric distribution without strong outliers, and the median with IQR for a skewed distribution or one with outliers. These pairings match each center with a spread measure that responds similarly to extreme values.
This pairing is a guide for describing distributions, not a requirement that every problem use both measures. If a question asks for the mean, calculate and report the mean even when the distribution is skewed; then explain how the shape affects its usefulness as a typical value. Similarly, a median remains a valid calculation for symmetric data. The decision is about which measure communicates the center most helpfully for the purpose at hand.
When comparing groups, use the same center measure when that makes sense and describe the groups in context. If one group is roughly symmetric and another has an extreme observation, explain why the most useful center may differ rather than treating the numbers as directly interchangeable.
Common Mistakes and AP Exam Tips
- Choosing the mean just because it is familiar. First inspect the shape and unusual values. A right-skewed distribution or a strong high outlier can pull the mean above the center of most observations.
- Claiming the mean is incorrect when there is an outlier. The mean is still correctly calculated. Explain instead that the extreme value makes it less representative of a typical observation.
- Assuming every skewed distribution must have an outlier. Skew describes an uneven tail; an outlier is an observation that stands apart. A distribution can be skewed without a single isolated value, as in the ticket-time example.
- Saying the median is unaffected by any change. The median is resistant to a few extreme values, but it can change if observations are added, removed, or changed in ways that alter the middle position or middle values.
- Reporting a center without units or context. “The median is 21” is incomplete if the reader does not know what was measured. State, for example, “The median appointment wait among the nine recorded appointments is 21 minutes.”
- Giving a choice without evidence. A full-credit explanation links the choice to a visible feature: “The distribution is right-skewed with a very long wait, so the median is more representative because the extreme wait pulls the mean upward.”
For full-credit communication, name the group and variable, identify the selected measure, give the value with units, and explain how the shape or outlier supports the choice. Avoid saying that a measure is “better” without specifying the purpose: the median may better describe a typical observation, while the mean still accounts for every value and describes the arithmetic balance point.
Check Your Understanding
For each situation, choose a measure of center and explain your choice using the distribution’s shape or unusual values.
- A set of twelve package weights has a roughly symmetric histogram and no observations that stand apart. Would the mean or median be a sensible primary description of center? Explain.
- Eight neighborhood walking times are close together, but one recorded time is much larger than the others. Which measure is likely to be more representative of a typical time, and why?
- A right-skewed distribution has a long tail but no single observation isolated from the rest. Is the median still a reasonable choice? Explain the difference between skew and an outlier.
- In the appointment-wait example, why did the mean change more than the median when the \(120\)-minute wait was included?
- Write a contextual center statement for the technical support ticket example using the measure selected in that example.