Statistics › Section B · Descriptive statistics › Topic B.5
Mean, median, and mode
Where do the data settle? The mean, the median and the mode answer in three different ways - and knowing which one to report is a judgement about the shape of the data, not a matter of taste.
Three ways to be typical
Where do the data settle? That question - the centre - has three standard answers. The mean adds and divides; the median locates the middle of the sorted list; the mode names the most frequent value. Each suits different data: the mean for symmetric quantitative data, the median for skewed data or data with outliers, and the mode the only option for nominal (unordered) categories, where an "average blood type" can only mean the most common one. This topic computes all three and builds the habit of pairing any centre with a measure of spread.
The mean
The mean is the sum of the values divided by how many there are:
It is the balance point of the data: values above it pull up exactly as hard as values below it pull down, which is why the deviations from the mean always total zero. Two consequences follow from the formula. Every value counts, so a single extreme value - a star athlete's salary in a ten-person firm - drags the mean toward itself. And the sample mean \(\bar{x}\) estimates the population mean \(\mu\); the law of large numbers says that larger and larger samples are very likely to land their means closer to \(\mu\).
The median
The median is the middle value once the data are sorted: half the values lie at or below it, half at or above - the same median met as Q2 in the percentile topic. For an odd count, the median is the single middle value; for an even count, it is the average of the two middle values. Because the median cares only about order, not size, replacing a billionaire neighbour with a millionaire leaves the median of a street's incomes untouched while the mean falls. A measure that extreme values barely move is called resistant.
The mode
The mode is the most frequent value, and the only measure of centre available for nominal categorical data, where "most common answer" is the only sensible centre. Quantitative sets can have one mode, several (two makes the data bimodal - often two groups mixed, as the histogram topic's two-peak shape showed), or none at all when nothing repeats. The mode also need not sit anywhere near the middle: a distribution can have its mode at its smallest value. For a quick "what's typical", the mode answers directly - the most popular shoe size, the most ordered drink.
Worked example: three centres, one data set
Ten students report hours of TV per day: 1, 1, 2, 2, 2, 3, 3, 4, 5, 7. 1. Mean: \(\bar{x} = \frac{1+1+2+2+2+3+3+4+5+7}{10} = \frac{30}{10} = 3\) hours. 2. Median: \(n = 10\) is even, so average the 5th and 6th sorted values: \(\frac{2+3}{2} = 2.5\) hours. 3. Mode: 2 hours, which appears three times. 4. Check: the mean sits above the median, exactly what the high tail (the 5 and the 7) should cause - one viewer's long hours pull the average above the typical viewer.
Worked example: the mean under attack
Return to the TV hours and change one value: the 7 becomes 30. The data are now 1, 1, 2, 2, 2, 3, 3, 4, 5, 30. 1. New mean: \(\frac{53}{10} = 5.3\) hours - the mean jumped from 3 to 5.3, although nine of the ten values never moved. 2. New median: still \(\frac{2+3}{2} = 2.5\) hours, because 2 and 3 remain the two middle values. 3. Check: one extreme watcher moved the mean by 2.3 hours and the median not at all. This resistance is why skewed data are summarised with the median.
Worked example: the mean from a frequency table
Raw lists are often unavailable; counts per value are. Six households own no pets, nine own one, four own two, and one owns three. 1. Total households: \(6+9+4+1 = 20\). 2. Sum of pets: \(0 \times 6 + 1 \times 9 + 2 \times 4 + 3 \times 1 = 0+9+8+3 = 20\). 3. Mean: \(20/20 = 1\) pet per household. 4. Check: nine of the twenty households own one pet, so a mean near one makes sense. Each value is weighted by its frequency - a shortcut that reproduces exactly the mean of the full list.
Grouped data: the midpoint estimate
When data arrive grouped into classes, the exact values are gone, so the mean can only be estimated: use each class midpoint as a stand-in for every value in the class. The coffee waits from the histogram topic were 18, 15, 11, 8, 5 and 3 customers in six two-minute bins, with midpoints 1, 3, 5, 7, 9 and 11 minutes:
The estimate is honest only if values sit roughly evenly inside their classes - one more reason to look at the shape before trusting any average.
Choosing the centre
Which centre to report is a judgement about shape. For roughly symmetric quantitative data, the mean is the standard choice: every value contributes, and mean, median and mode land close together. For skewed data or data with outliers, the median is the honest centre - the TV example showed one long-hours watcher dragging the mean half an hour above the median. For nominal categorical data, the mode is the only option. Whatever the choice, report a centre alongside a measure of spread - two cities can share a mean temperature while one swings between extremes - and for skewed data the median pairs naturally with the IQR.
Common mistakes to avoid
First, averaging ordered categories as if they were numbers - an "average satisfaction rating" invents precision the scale never had. Second, quoting a mean for heavily skewed data without the median beside it. Third, reporting a centre with no spread measure at all. Fourth, taking one of the two middle values as the median when \(n\) is even instead of averaging them. Fifth, treating the midpoint estimate from grouped data as exact. Sixth, assuming every data set has a mode: when nothing repeats, there is none.
Key takeaways
The mean \(\bar{x} = \frac{\sum x_i}{n}\) is the balance point, uses every value, and is pulled by extremes. The median - the middle of the sorted data, averaging the two middle values when \(n\) is even - is resistant to them. The mode is the most frequent value and the only centre for nominal categorical data. From a frequency table, weight values by their counts; from grouped data, estimate with class midpoints. Report the mean for symmetric data, the median for skewed data or data with outliers, and always pair the centre with a measure of spread.
Flashcards
Flashcard
Sum of the values divided by n
Test your understanding
Question 1 of 5
What is the mean of 2, 4, 6 and 8?
Question 2 of 5
What is the median of 3, 10, 4 and 7?
Question 3 of 5
Household incomes in one town are heavily right-skewed. Which centre should the town report?
Question 4 of 5
The data 2, 2, 3, 3, 5 have what mode or modes?
Question 5 of 5
In a survey, 5 households own no cars, 8 own one, 5 own two and 2 own three. What is the mean number of cars?
