Statistics › Section B · Descriptive statistics › Topic B.3

Percentiles and quartiles

Knowing a value tells you little until you know where it sits among its fellows. Percentiles measure relative standing - and their quartile cousins split data into quarters and hand us a spread measure that outliers barely disturb.

Where do you stand?

Is a 6-minute wait long? Compared with what? Nearly every "is this good?" question about a number is really a question about position within a group, and percentiles are the standard answer. Colleges read SAT scores as percentiles, doctors read a child's height as a percentile on a growth chart, and marathon organisers publish finishing-time percentiles. This topic defines percentiles, builds the quartiles - special percentiles that split data into quarters - and develops the interquartile range, a spread measure that extreme values barely move.

What a percentile means

The kth percentile of an ordered data set is a value with roughly k percent of the data at or below it and the rest above. Being at the 90th percentile on a test does not mean you answered 90 percent correctly - it means 90 percent of scores fell below yours. Percentiles compare rather than score, so the same value can sit at different percentiles in different groups. And direction depends on the variable: a high percentile is good for savings or income, but for race times or hospital waits a low percentile is better. Always read a percentile in context.

Finding the kth percentile

To find the kth percentile from ordered data: 1. Count the values, \(n\). 2. Compute the position \(i = \frac{k}{100}(n + 1)\). 3. If \(i\) is a whole number, the kth percentile is the value in position \(i\), counting from the smallest. 4. If \(i\) is not whole, round it down and up to the two neighbouring positions and average the values there. With 15 quiz scores, the 30th percentile gives \(i = 0.30 \times 16 = 4.8\): positions 4 and 5 hold the values 6 and 7, so the percentile is \(\frac{6+7}{2} = 6.5\). The method always starts the same way: sort the data.

Worked example: quartiles of bakery sales

A bakery records pastry sales, in dozens, on 20 days, sorted: 12, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 31, 33, 36, 55. Find Q1, the median and Q3. 1. \(n = 20\), so \(n + 1 = 21\). 2. Q1, the 25th percentile: \(i = 0.25 \times 21 = 5.25\), so average the 5th and 6th values: \(\frac{18+19}{2} = 18.5\). 3. Median, the 50th: \(i = 10.5\), so average the 10th and 11th: \(\frac{23+24}{2} = 23.5\). 4. Q3, the 75th: \(i = 15.75\), so average the 15th and 16th: \(\frac{28+29}{2} = 28.5\). 5. Check: five values sit below 18.5 (a quarter of 20), ten below 23.5 and fifteen below 28.5 - the positions match the meanings.

Finding a value's percentile rank

The reverse question - what percentile does this value sit at? - is answered by counting. Let \(x\) be the number of values below it and \(y\) the number equal to it. Its percentile rank is

\frac{x + 0.5y}{n} \times 100

Counting half of the tied values on each side splits ties fairly. Percentile ranks are how "top ten percent" claims are checked, and how you can locate any score in its distribution once you know the counts.

Worked example: a score's percentile

Fifteen students score on a 10-point quiz, sorted: 4, 5, 6, 6, 7, 7, 7, 8, 8, 8, 8, 9, 9, 10, 10. Where does a score of 8 stand? 1. Values below 8: \(x = 7\). 2. Values equal to 8: \(y = 4\). 3. Percentile rank: \(\frac{7 + 0.5 \times 4}{15} \times 100 = \frac{9}{15} \times 100 = 60\). A score of 8 sits at the 60th percentile: counting half the ties, 60 percent of the class scored below it. 4. Check: 7 below plus half of the 4 ties is 9 of 15 values, and \(9/15 = 0.60\), so the rank is confirmed.

Quartiles and the interquartile range

Three quartiles split ordered data into quarters. Q1, the first quartile, is the 25th percentile; Q2 is the median, the 50th; Q3 is the 75th. The interquartile range,

\text{IQR} = Q_3 - Q_1

measures the spread of the middle half of the data - the distance across the central 50 percent of values. Because the IQR ignores the lowest and highest quarters, making the largest value far larger leaves it unchanged, which makes it the natural companion to percentiles whenever outliers are in play.

Worked example: the 1.5 × IQR rule

Which bakery sales count as unusual? From the sorted sales, \(Q_1 = 18.5\) and \(Q_3 = 28.5\), so \(\text{IQR} = 10\). 1. Lower fence: \(Q_1 - 1.5 \times \text{IQR} = 18.5 - 15 = 3.5\). 2. Upper fence: \(Q_3 + 1.5 \times \text{IQR} = 28.5 + 15 = 43.5\). 3. Compare the data: every value from 12 to 36 sits inside the fences, but 55 exceeds 43.5 - a potential outlier. 4. Check: the fences sit one and a half IQR-widths outside each quartile, and 55 lies well beyond that distance above Q3. A flagged value deserves investigation, never automatic deletion.

Percentiles from grouped tables

When data arrive already grouped in a frequency table, percentiles come from the cumulative relative frequency column. Scan down the cumulative column until it first reaches the target share; the value in that row is the percentile. Fifty quiz scores: 2 students scored 5, 5 scored 6, 8 scored 7, 14 scored 8, 12 scored 9 and 9 scored 10, giving cumulative shares 0.04, 0.14, 0.30, 0.58, 0.82 and 1.00. The 80th percentile is a score of 9, because the cumulative share first reaches 0.80 in the row for 9. Grouped answers are coarser, so treat them as approximations.

Common mistakes to avoid

First, reading a percentile as a percentage correct: the 90th percentile is about position, not points. Second, computing percentiles on unsorted data - every rule in this topic starts with sort. Third, forgetting the tie-splitting half in the rank formula, which shifts tied values by several places. Fourth, calling any large value an outlier without the fence calculation, or deleting a flagged value on the spot. Fifth, using \(n\) where the rule says \(n + 1\), or mixing methods: software and other books use slightly different rules, so state your method whenever you report a percentile.

Key takeaways

The kth percentile has roughly k percent of the ordered data at or below it - a measure of relative standing, always read in context. Find it with \(i = \frac{k}{100}(n+1)\): the exact position when \(i\) is whole, otherwise the average of the two neighbouring positions. A value's own percentile rank is \(\frac{x + 0.5y}{n} \times 100\), with \(x\) values below and \(y\) tied. The quartiles Q1, median (Q2) and Q3 mark the 25th, 50th and 75th percentiles, and \(\text{IQR} = Q_3 - Q_1\) spreads the middle half, resistant to extremes. Values outside \(Q_1 - 1.5\,\text{IQR}\) or \(Q_3 + 1.5\,\text{IQR}\) are potential outliers to investigate.

Flashcards

Flashcard

kth percentile

A value with roughly k percent of the ordered data at or below it

Position within a group, not a score.

Test your understanding

Question 1 of 5

Your test score is at the 90th percentile. This means:

Question 2 of 5

In 19 ordered values, i for the 25th percentile is 0.25 × (19 + 1) = 5, a whole number. So Q1 is:

Question 3 of 5

In the quiz data, the score 8 has 7 scores below it and 4 tied with it, with n = 15. Its percentile rank is:

Question 4 of 5

With Q1 = 18 and Q3 = 28, what is the IQR?

Question 5 of 5

With Q1 = 18, Q3 = 28 and IQR 10, which values does the 1.5 × IQR rule flag as potential outliers?

← Previous topicNext topic →