How to find a percentile
The p-th percentile, Pₚ, is the value with p per cent of the list at or below it. There is no formula to put numbers into, only a procedure. This page uses one common in school textbooks:
- Sort the n values.
- Work out the position k = p/100 · n.
- If k is a whole number, the percentile is the mean of value number k and value number k + 1.
- If k is not a whole number, round it up and take that value.
The position is not the answer. It only says where in the list the answer is.
Worked example
Fifteen test scores, already sorted:
12, 15, 18, 20, 21, 23, 25, 26, 28, 30, 33, 35, 38, 41, 47
P₉₀. The position is 0.9 · 15 = 13.5. Not a whole number, so round up to 14. Value number 14 is P₉₀ = 41. Check: 14 of the 15 scores, 93%, are at or below 41, and 2 of 15 are at or above it.
P₂₀. The position is 0.2 · 15 = 3. A whole number, so take the mean of values 3 and 4: (18 + 20)/2 = P₂₀ = 19.
P₅₀. The position is 7.5, up to 8, and value 8 is 26 — the same number as the median. It should be: the 50th percentile is the median, whatever the count.
Two mistakes that give plausible numbers
Reading the percentile as a share of the largest value. 90% of 47 is 42.3, which is not even in the list. A percentile is a value taken from the list, chosen by how many lie below it.
Rounding the position down. 13.5 down is 13, and value 13 is 38. But only 13 of 15, 87%, are at or below 38 — too few for the 90th percentile. That is why it rounds up.
Why a spreadsheet gives another answer
Type the same list into a spreadsheet and ask for PERCENTILE.INC(…, 0.9): the answer is 39.8. The
spreadsheet puts the position at (n − 1) · p/100 + 1 = 13.6 and goes 0.6 of the way from value 13 to
value 14. That is another definition, not an arithmetic slip — statistics software knows about nine —
but on a test the one that counts is the one the course teaches.
| Method | P₂₀ | P₉₀ |
|---|---|---|
| Position p/100 · n, round up or average (this page) | 19 | 41 |
Spreadsheet PERCENTILE.INC / PERCENTILE |
19.6 | 39.8 |
A second example: the other way round
Often the question runs backwards: you have a score and want to know which percentile it is on. A score of 33 in the list above is in position 11. Eleven of fifteen, 73%, scored 33 or less, so the score is at about the 73rd percentile. That says nothing about how much of the test was right — only how the score stands against the others.
Where next
The 25th, 50th and 75th percentiles have names of their own: the quartiles. The distance from the first to the third is the interquartile range, and a box plot draws them with the median and the extremes.
Common mistakes
- Not sorting first. Position 14 in an unsorted list is just whatever value happens to sit there.
- Answering with the position. 13.5 or 14 is where the answer is, not the answer.
- Rounding the position down. It lands one value too low, with too few values below it.
- Taking p per cent of the largest value. A percentile is about how many, not how much.
- Mixing methods. A spreadsheet and a textbook can give different answers for one list. Say which method you used.
Common questions
Are P₂₅ and P₇₅ the same as the quartiles?
Yes, the lower and upper quartile. With the method on this page they usually come out the same as
the median of each half. The exception is a count one more than a multiple of four, such as 9 or 13:
there the halves give the mean of two values and the position rule gives one of them. For quartiles,
use the halves.
What does it mean that a child is on the 75th percentile for height?
That 75% of children of the same age are as tall or shorter, and 25% taller. It is a place in the
group, not a grade: the 10th and the 90th percentile are both normal heights.
Why do different calculators give different percentiles?
Because there is no single agreed definition. On long lists the difference is small; on short ones it
can be a whole step. That is one reason percentiles are mostly used on large data sets.