Three averages, three different questions
“Average” is an everyday word for a typical value, and there is more than one way to pick it. The mean, median and mode each answer a slightly different question, and they agree only when the data is symmetric.
Mean = Σx ÷ n
Median = the middle value of the sorted data (n odd), or the mean of the two middle values (n even)
Mode = the value that occurs most often Range = largest value − smallest value
Here x stands for each value, Σx is the total of all the values and n is how many values there are. In a sorted list of n values the median sits at position (n + 1) ÷ 2; when that is a half-number such as 4.5, average the values either side of it.
Worked example by hand
Seven people are asked how many kilometres they travel to work: 5, 9, 3, 34, 6, 5, 8.
- Sort the data. 3, 5, 5, 6, 8, 9, 34.
- Mean. The total is 70 and there are 7 values, so the mean is 70 ÷ 7 = 10.
- Median. The middle position is (7 + 1) ÷ 2 = 4, and the 4th value in order is 6.
- Mode. 5 appears twice and everything else once, so the mode is 5.
- Range. 34 − 3 = 31.
The mean of 10 is larger than six of the seven values. One long commute has dragged it upwards, while the median of 6 still describes the typical person. That is the whole reason for having more than one average.
With an even count there is no single middle value. For 3, 5, 6, 8 the middle two are 5 and 6, so the median is 5.5, a number that does not appear in the data at all.
Which average suits which data
Mean: symmetric measurements without extreme values
The mean uses every value, which makes it the most informative average when the data is roughly symmetric: heights, test scores, repeated measurements of the same thing. It is also the only one of the three that lets you get back to the total, since mean × n = sum. Its weakness is that every value pulls on it in proportion to its distance, so one extreme figure can move it a long way.
Median: skewed data and data with outliers
Incomes, house prices, waiting times and anything else with a long tail on one side are better summarised by the median. Half the values lie below it and half above, whatever the extremes do. Replace the 34 in the example with 340 and the mean jumps to 53.71, while the median stays at 6. This is why property reports quote median prices.
Mode: categories and “most popular” questions
The mode is the only average that works for data that is not numerical, such as shoe sizes stocked, favourite colours or the most common answer in a survey. For measurements with many decimal places it is rarely useful, because values seldom repeat exactly.
No mode and more than one mode
If every value occurs exactly once, no value is more common than any other and the data has no mode. The calculator says “None” in that case. If two or more values tie for the highest count, they are all modes: 2, 2, 5, 7, 7 has the two modes 2 and 7 and is called bimodal. Two clear modes often mean that two different groups have been mixed together, which is worth investigating before quoting any single average.
Geometric and harmonic means
Geometric mean = (x₁ × x₂ × … × xₙ)1 ÷ n Harmonic mean = n ÷ (1 ÷ x₁ + 1 ÷ x₂ + … + 1 ÷ xₙ)
Both need every value to be greater than zero, and the calculator shows a dash otherwise.
Use the geometric mean for things that multiply, such as growth rates. An investment that returns +10%, +20% and −15% in successive years has growth factors 1.10, 1.20 and 0.85. Their product is 1.122, and the cube root of that is 1.0391, so the average growth is 3.91% a year. The ordinary mean of the three returns is 5%, which overstates it: three years at 5% would give a factor of 1.1576, not 1.122.
Use the harmonic mean for rates over equal distances or equal amounts of work. Drive out at 60 km/h and back along the same road at 40 km/h and your average speed is 2 ÷ (1 ÷ 60 + 1 ÷ 40) = 48 km/h, not 50, because you spend longer at the slower speed.
For any set of positive numbers that are not all equal, the mean is the largest of the three, the geometric mean comes next and the harmonic mean is the smallest. For the commuting data they are 10, 7.30 and 6.01.
Common mistakes
- Forgetting to sort before finding the median. The middle of the list as typed, 34 in the example, is not the median.
- Averaging averages. The mean of two group means is the overall mean only if the groups are the same size. Otherwise, add up the totals and divide by the combined count.
- Quoting the mode’s frequency as the mode. In the example the mode is 5, the value, not 2, the number of times it appears.
- Treating the range as a measure of typical spread. It depends only on the two most extreme values. The standard deviation or the interquartile range from the statistics calculator says more.