What a Percentile Rank Measures
A percentile rank tells us how a specified value compares with the observations in a data set. When some observations are tied at that value, the calculation must say what happens to those ties. In this tutorial, the rule is direct: count every observation at or below the specified value, then express that count as a percentage of all observations.
For example, a score’s percentile rank is not another score. It is a percentage describing the score’s position among the observations. Ties matter because every observation equal to the specified score is included in the count. This gives a clear, repeatable answer even when several observations share the same value.
This definition is a descriptive calculation on the data set at hand. It does not require a distributional model or an assumption that the data follow a particular shape. The key is to identify the full set of observations being described, specify the comparison value, and count correctly.
A Counting Formula That Handles Ties
It can help to split the count into two parts. Let \(B\) be the number of observations strictly below \(x\), and let \(T\) be the number of observations equal to \(x\). The number at or below \(x\) is \(B+T\). Thus:
This formula makes the treatment of ties explicit. If three observations equal \(x\), all three are part of \(T\) and all three contribute to the percentile rank. Counting only values strictly below \(x\) would instead give a different percentage, which answers a different question.
Worked Example: A Score Tied Three Times
Suppose a hypothetical group of 12 students receives these quiz scores, listed in order: 52, 60, 68, 72, 75, 80, 80, 80, 84, 88, 91, and 96. Find the percentile rank of a score of 80, using the percentage at or below the score.
There are \(n=12\) scores. Five scores are strictly below 80: 52, 60, 68, 72, and 75. Three scores equal 80, so \(B=5\) and \(T=3\). Therefore, \(B+T=8\) scores are at or below 80.
The score of 80 has a percentile rank of about 66.7% in this group: 66.7% of the 12 scores are 80 or lower. The three tied scores are all included. If we counted only scores strictly below 80, the result would be \(\frac{5}{12}\times100\%\approx41.7\%\), which does not answer the at-or-below question.
Calculation From a Frequency Table
A frequency table groups equal values together and reports how many observations have each value. The frequency at the specified value represents the ties, so it must be added to the frequencies of all smaller values. The total number of observations is the sum of every frequency in the table.
Worked Example: Counting Ties in a Frequency Table
A hypothetical set of 20 laboratory practice scores is summarized below. Find the percentile rank of 7 points.
| Score | Frequency |
|---|---|
| 4 | 2 |
| 5 | 3 |
| 6 | 4 |
| 7 | 5 |
| 8 | 4 |
| 9 | 2 |
The total frequency is \(2+3+4+5+4+2=20\), so \(n=20\). The number below 7 is \(2+3+4=9\). The frequency at 7 is 5, meaning five observations are tied at the specified value. Thus, 14 observations are at or below 7.
A score of 7 has a percentile rank of 70% in this group: 70% of the practice scores are 7 points or lower. The four observations with scores of 8 and the two with scores of 9 are not included because they exceed 7.
Why the Calculation Works
The calculation is a proportion expressed as a percentage. Each observation either meets the condition “at or below \(x\)” or does not. Assign an indicator of 1 to each observation that meets the condition and 0 to each that does not. The sum of these indicators is exactly the count \(C\). Dividing by \(n\) gives the fraction of observations that meet the condition; multiplying by 100 converts that fraction to a percentage.
Here \(x_i\) is the \(i\)th observation, and \(I(x_i\leq x)\) equals 1 when \(x_i\leq x\) and 0 otherwise. If \(T\) observations equal \(x\), each of those has \(x_i\leq x\), so each contributes 1 to the sum. That is why ties are automatically included by the at-or-below condition.
The result must be between 0% and 100%, inclusive. No observations at or below \(x\) gives 0%; all observations at or below \(x\) gives 100%. As the comparison value increases, the count at or below it cannot decrease, so the percentile rank cannot decrease either. These checks can help catch a counting or arithmetic error.
Worked Example: A Value Between Observed Scores
Suppose a hypothetical set of 10 waiting times, in minutes, is 4, 5, 5, 6, 7, 7, 7, 8, 10, and 12. Find the percentile rank of 7.5 minutes.
Although 7.5 is not an observed waiting time, the definition still applies. There are seven observations at or below 7.5: 4, 5, 5, 6, 7, 7, and 7. There are no observations equal to 7.5, so the count is simply the number below it. The total is \(n=10\).
The percentile rank of 7.5 minutes is 70%: 70% of these waiting times are at or below 7.5 minutes. The waiting-time unit describes the comparison value, while the result itself is a percentage and therefore has no time unit.
Percentile Rank Is Not a Percentile Value
The terms are related but answer different questions. A percentile rank starts with a value and reports the percentage of observations at or below it. A percentile value starts with a target percentage and asks for a value associated with that position in the ordered data. One is a percentage calculated from a value; the other is a value associated with a percentage.
| Question | What is specified? | What is found? |
|---|---|---|
| What is the percentile rank of a score of 80? | A value: 80 | The percentage of observations at or below 80 |
| What score corresponds to a chosen percentile? | A target percentage | A value in the data scale |
For example, “80 has a percentile rank of about 66.7%” means that about 66.7% of the observations are at or below 80. It does not mean that 80 is the 66.7th percentile under every method for selecting percentile values. Different procedures for finding percentile values may handle ranks and ties differently; do not use a percentile-rank calculation as a substitute for such a procedure.
Common Mistakes and Checks
The most common error is to count only observations strictly below the specified value. That omits the tied observations, even though the definition asks for the percentage at or below the value. A second error is to report the count itself as the percentile rank: the count must be divided by the total number of observations and converted to a percentage.
Before reporting a result, check that the numerator counts observations rather than distinct values. In a data set with five repeated scores, those are five observations, not one. Also check that the denominator includes all observations in the group being described. Finally, state the interpretation in context: name the comparison value, the group, and the phrase “at or below.”
- Identify the specified value and the data set whose percentile rank is being calculated.
- Count every observation below the value and every observation tied at it.
- Divide that combined count by the total number of observations and multiply by 100%.
- Describe the result as a percentage at or below the specified value, not as a data value.
Check Your Understanding
Answer each question using the at-or-below definition of percentile rank.
- For the scores 3, 4, 4, 6, 6, 6, 8, and 9, find the percentile rank of 6. Show how ties enter the calculation.
- In a group of 25 observations, 11 are below a value and 4 equal it. What is the percentile rank of that value?
- A value is not present in a data set. What observations should be counted to find its percentile rank?
- Explain why the count of values strictly below \(x\) alone does not give the percentile rank defined in this tutorial when there are ties at \(x\).
- In your own words, distinguish the question answered by a percentile rank from the question answered by a percentile value.