Five numbers drawn as a picture. A box plot is built from the minimum, the first quartile, the median, the third quartile and the maximum, with a box spanning the middle fifty percent and whiskers reaching the extremes; because each of the four parts holds about a quarter of the data, the width of a part reports how tightly that quarter is packed rather than how many values it contains. Then the modified box plot, whose whiskers stop at the most extreme non-outliers and whose outliers are marked as separate points, and parallel box plots on one shared number line, which are the cheapest honest comparison of several distributions. Closes with what a box plot cannot show.
Subject: Statistics · 66 slides · symbolic lesson
Open the interactive version of this deck
Title
Statistics · Chapter 2 — Descriptive Statistics
Five numbers drawn to scale: where the data crowd, where they thin out, and how groups compare
Objectives
Section 2.3 gave you quartiles: numbers that cut a sorted data set into four groups of about the same size. This section turns five of those location numbers into a picture you can read in two seconds, and that you can stack beside other pictures to compare groups.
The arithmetic in this lesson is light: sorting, averaging two numbers, and subtracting. What takes practice is the reading. Almost everybody, on first meeting, reads a long whisker as meaning lots of data. By the end of Part 3 you will read it the right way without thinking.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 94-98 — the section this deck covers, with Example 2.22 from p. 94
Warm-up
Everything in this lesson rests on one sentence from Section 2.3: a quartile is a cut with about a quarter of the data on one side. Try It 2.19 asks you to say that sentence in context.
Discussion prompt
For the 100-meter dash, the third quartile of finishing times was 11.5 seconds. Interpret the third quartile in the context of the race.
Notice what the answer is about: a share of the runners, not a share of the time. A quartile is defined by counting people, and that is the idea this whole lesson keeps returning to. Hold on to it when the picture tempts you to measure instead.
Also recall how the book finds a quartile by hand: sort the data, find the median, and then take the median of the lower half for Q1 and of the upper half for Q3. When the count is odd, the median itself is left out of both halves.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 93 — Try It 2.19
Section
Part 1
Concept
Example 2.23 lists the heights of forty students in a statistics class, in inches. Here they are, already sorted. Read them for ten seconds and then try to describe the class.
\[ 59, 60, 61, 62, 62, 63, 63, 64, 64, 64, 65, 65, 65, 65, 65, 65, 65, 65, 65, 66, \]
\[ 66, 67, 67, 68, 68, 69, 70, 70, 70, 70, 70, 71, 71, 72, 72, 73, 74, 74, 75, 77 \]
Most people manage something like around sixty-five and stop. Forty numbers do not fit in one look, and the list hides its own shape: you cannot see at a glance where the heights crowd together or how far the tallest few stretch.
The book's answer is to keep just five numbers, chosen so that between them they mark off four groups of ten students. Those five numbers locate the ends of the data and the three quartile cuts.
five-number summary — The minimum, the first quartile Q1, the median, the third quartile Q3 and the maximum of a data set, listed in that order. They cut the sorted data into four parts holding about a quarter of the values each.
\[ \text{min}, \quad Q_1, \quad \text{median}, \quad Q_3, \quad \text{max} \]
Five numbers will never tell you everything the forty did. The rest of this lesson is about what they do tell you, how to draw them, and what they leave out.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 94-95 — the five values a box plot is built from
Concept
Every number in the summary is found by counting positions in the sorted list, never by measuring along the scale. Sort first, then locate the median by its place.
\[ \text{median place} = \frac{n + 1}{2} \]
With forty values that place is 20.5, which means halfway between the 20th and 21st values. With fifteen values it is the 8th value exactly. A place ending in .5 always means average the two neighbours.
Then split the sorted list at the median into a lower half and an upper half of equal size. Q1 is the median of the lower half and Q3 is the median of the upper half. When n is odd, the median value belongs to neither half, so each half holds the remaining values split evenly.
\[ n = 40: \quad \text{halves of } 20, \quad Q_1 \text{ at place } 10.5, \quad Q_3 \text{ at place } 30.5 \]
\[ n = 15: \quad \text{median at place } 8, \quad \text{halves of } 7, \quad Q_1 \text{ at place } 4, \quad Q_3 \text{ at place } 12 \]
Calculators and software sometimes use slightly different quartile rules, so a computer may report a Q1 that differs a little from yours on small data sets. This deck uses the book's rule throughout, and every quartile here was rechecked with it.
OpenStax Introductory Business Statistics 2e, §2.2 Measures of the Location of the Data §2.2 Measures of the Location of the Data — the same median-of-halves rule
Notation
Two differences of the five numbers have names, and both appear on almost every box plot question. Hover over each piece; notice which numbers each spread uses and which it ignores.
Annotate
On: \( \text{range} = \text{max} - \text{min}, \qquad IQR = Q_3 - Q_1 \)
Read them as two different questions. The range asks how far apart the most extreme values are. The IQR asks how spread out the typical half of the data is. A data set can have a huge range and a tiny IQR, which means a tight middle with one or two values far away.
Both spreads are measured in the units of the data, so always write the unit: a range of 18 inches, an IQR of 40 minutes. A spread without units cannot be compared with anything.
On a box plot you can see both at once: the range is the length of the whole drawing, whisker tip to whisker tip, and the IQR is the length of the box alone.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Example 2.23(c)-(d), the range and the interquartile range
Worked example
The book's first box plot is built from fourteen values, already in order. Find the five-number summary.
\[ 1, 1, 2, 2, 4, 6, 6.8, 7.2, 8, 8.3, 9, 10, 10, 11.5 \]
Fourteen is even, so the median will fall between two values and each half will hold exactly seven. Expect Q1 and Q3 to be actual data values, since the middle of seven is a single place.
Read the extremes
Why: the first and last values of the sorted list.
\[ \text{min} = 1, \qquad \text{max} = 11.5 \]
Locate the median place
Why: fourteen values.
\[ \frac{14 + 1}{2} = 7.5 \;\Rightarrow\; \text{average the 7th and 8th} \]
Average them
Why: the 7th value is 6.8 and the 8th is 7.2.
\[ \text{median} = \frac{6.8 + 7.2}{2} = \frac{14}{2} = 7 \]
Q1 is the middle of the lower seven
Why: places 1 to 7; the middle is the 4th.
\[ 1, 1, 2, \underline{2}, 4, 6, 6.8 \;\Rightarrow\; Q_1 = 2 \]
Q3 is the middle of the upper seven
Why: places 8 to 14; the middle is the 11th.
\[ 7.2, 8, 8.3, \underline{9}, 10, 10, 11.5 \;\Rightarrow\; Q_3 = 9 \]
Figure (svg): A box plot of the fourteen values 1, 1, 2, 2, 4, 6, 6.8, 7.2, 8, 8.3, 9, 10, 10 and 11.5 on a number line from 0 to 12. The box runs from 2 to 9 with a yellow median line at 7, and whiskers reach out to 1 and to 11.5. Labels above name the five numbers: min 1, Q1 = 2, median 7, Q3 = 9, max 11.5. Underneath, orange dots mark the fourteen values themselves.
In context: the middle half of these fourteen values lies between 2 and 9, an IQR of 7, while the whole set spans a range of 10.5. The median, 7, is a value nobody in the list actually has, which is normal when n is even.
Check the order and the shares
Why: the five must not decrease, and each cut leaves about a quarter on each side.
\[ 1 \le 2 \le 7 \le 9 \le 11.5, \quad \text{and 7 values lie on each side of } 7 \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 95 — the dataset of Figure 2.11
Worked example
Find the five-number summary of the forty heights. The book gets these from a calculator; here they come from counting places, so you can do it with a pencil.
The plan is to locate three places, 20.5, 10.5 and 30.5, and read the values there. Because forty is even, every one of those places falls between two students, so each answer is an average of two neighbours.
Read the extremes
Why: first and last of the sorted list.
\[ \text{min} = 59, \qquad \text{max} = 77 \]
Median place
Why: forty values.
\[ \frac{40 + 1}{2} = 20.5 \;\Rightarrow\; \text{average places 20 and 21} \]
Both places hold 66
Why: average them.
\[ \text{median} = \frac{66 + 66}{2} = 66 \]
Q1: the lower twenty, places 1 to 20
Why: its middle is between places 10 and 11.
\[ Q_1 = \frac{64 + 65}{2} = 64.5 \]
Q3: the upper twenty, places 21 to 40
Why: its middle is between places 30 and 31.
\[ Q_3 = \frac{70 + 70}{2} = 70 \]
Figure (svg): The forty heights in order, in two rows of twenty cells, with small grey place numbers above. Places 1 to 10 are blue, 11 to 20 orange, 21 to 30 purple and 31 to 40 green. A yellow cut between places 10 and 11 is labelled Q1, the average of 64 and 65, which is 64.5. A pink cut at the end of the first row is labelled median, the average of places 20 and 21, both 66. A yellow cut between places 30 and 31 is labelled Q3, the average of 70 and 70, which is 70.
In context: the middle half of the class stands between 64.5 and 70 inches, and no student is exactly 64.5 inches tall. A quartile is a cut between people, not necessarily somebody's height.
Check against the book
Why: the calculator values printed on p. 95.
\[ 59, \; 64.5, \; 66, \; 70, \; 77 \quad \text{matches Example 2.23} \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 95 — Example 2.23
Fill the middle
The page counts of forty books, sorted, are below. Places 10 and 11 hold 234 and 240; places 20 and 21 hold 318 and 326; places 30 and 31 hold 392 and 398. Fill in the three cuts and the interquartile range.
\[ 136, 140, 178, \dots, 234, 240, \dots, 318, 326, \dots, 392, 398, \dots, 475, 512 \]
Fill in the blanks
Q_1 = \frac237322 = 395, \quad M = \frac158___ = ___, \quad Q_3 = ___, \quad IQR = ___
Why: Each cut averages its two neighbours: 474 over 2 is 237, 644 over 2 is 322 and 790 over 2 is 395. The interquartile range is 395 minus 237, which is 158 pages.
Every one of these answers is a page count that no book on the shelf actually has, because no two neighbouring books tie. That is fine: the cuts sit between books, with ten books on each side of every cut.
The book asks you to state the interquartile range, and the answer needs its unit: the middle twenty books span 158 pages. Keep this summary, 136, 237, 322, 395 and 512; Part 2 draws it.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Try It 2.23
Prediction
Suppose the tallest student in Example 2.23 were 90 inches instead of 77. Every other height stays the same. Before you reason it through, commit to which of the summary's numbers change.
Think about where the tallest student sits in the sorted list, and which places each of the five numbers is read from. A number can only change if a value at one of its places changes.
Predict first
With the tallest height changed from 77 to 90, which numbers in the summary change?
Correct: Only the maximum, and so the range
Why: The tallest student is at place 40. Q3 depends only on places 30 and 31 and the median only on places 20 and 21, so neither moves. The maximum becomes 90 and the range grows from 18 to 31, while the IQR stays 5.5.
This is why the quartiles are called resistant. They are positions in the count, so an extreme value can only change them if it changes which values occupy the middle places, and moving the tallest student further up never does that.
The range, by contrast, is built from the two most extreme values alone. One unusual student can make it as large as you like. Keep that contrast in mind: Part 4 uses the resistant IQR to judge whether an extreme value is unusual.
Intuition
The prediction you just made is worth understanding rather than memorising, because Part 4 is built on it. Every quartile is read from fixed places in the sorted list. For forty values those places are 10 and 11, 20 and 21, and 30 and 31, whatever the values are.
\[ Q_1 \leftarrow \text{places 10, 11}, \quad M \leftarrow \text{places 20, 21}, \quad Q_3 \leftarrow \text{places 30, 31} \]
Now change the tallest student from 77 to 90, or to 900. That student stays at place 40, the last place, so the values sitting at places 30 and 31 do not change, and neither does Q3. The cut only moves if a value crosses it and changes which value sits at its place.
\[ \text{range: } 77 - 59 = 18 \;\longrightarrow\; 90 - 59 = 31, \qquad IQR: \; 5.5 \;\longrightarrow\; 5.5 \]
The range, by contrast, is made from places 1 and 40 alone, so it follows the extreme value wherever it goes. Statisticians call a summary like the IQR resistant: it resists being dragged by a few unusual values. A box plot is drawn mostly from resistant numbers, which is why its box stays put while its whiskers swing.
Remember the limit of the idea, though. Resistant does not mean unchangeable. If ten students suddenly grew, values would cross the cuts and the quartiles would move with them. One extreme value cannot move a quartile; a shift in the bulk of the data can.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Example 2.23(c)-(d), the range and the IQR
Trap
A first attempt finds Q1 by measuring along the scale instead of counting students:
\[ \frac{77 - 59}{4} = 4.5 \]
\[ Q_1 = 59 + 4.5 = 63.5 \]
Wrong. Only 7 heights lie below 63.5, not 10.
The arithmetic is fine; the idea is wrong. Dividing the range into four equal lengths gives equal pieces of the scale, and a quartile is not a piece of the scale.
A quartile cuts the count, not the scale. Count to the middle of the lower twenty places instead.
\[ Q_1 = \frac{64 + 65}{2} = 64.5 \]
The two methods agree only when the data are spread evenly over the scale. This class crowds around 65 inches, and crowding is exactly what measuring the scale cannot see.
Check any quartile by counting: ten of the forty heights lie below 64.5, a quarter of them, as a first quartile promises.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 95 — Example 2.23
Section
Part 2
Concept
The book's recipe is short. Use a horizontal or vertical number line, scaled so that it covers every data value. Then draw a box whose two ends sit at Q1 and Q3, a line inside the box at the median, and whiskers from the ends of the box out to the smallest and largest values.
Figure (svg): The fourteen-value box plot again, on a number line from 0 to 12, with coloured brackets under its four parts: lower whisker from 1 to 2, lower half of the box from 2 to 7, upper half of the box from 7 to 9, and upper whisker from 9 to 11.5. A dashed line above the box from 2 to 9 is labelled the box, about the middle 50 percent. A note says each bracket holds about a quarter of the values.
Each piece of the drawing is one of the five numbers. The left whisker ends at the minimum, the box starts at Q1, the line marks the median, the box ends at Q3, and the right whisker ends at the maximum. Nothing else is drawn, so a box plot carries exactly five numbers of information.
Because the quartiles cut the data into quarters, approximately the middle 50 percent of the data fall inside the box, and about a quarter of the data sit along each whisker. The median line can sit anywhere in the box, and it can even land on one of the box's ends.
box plot — A display of the five-number summary on a scaled number line: a box from Q1 to Q3 holding about the middle half of the data, a line at the median, and whiskers from the box out to the minimum and the maximum. Also called a box-and-whisker plot.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 94-95 — constructing a box plot and Figure 2.11
Worked example
Draw the box plot of the forty heights from their summary: 59, 64.5, 66, 70 and 77 inches.
Draw the axis before anything else. If you draw a box first and then try to fit numbers to it, the spacing will come out wrong, and in a box plot the spacing is the whole message.
Choose the scale
Why: it must cover 59 to 77; ticks every 2 inches from 58 to 78 are easy to read.
\[ \text{axis: } 58 \text{ to } 78, \quad \text{ticks every 2 inches} \]
Draw the box between the quartiles
Why: its width is the IQR.
\[ \text{box from } 64.5 \text{ to } 70, \quad IQR = 70 - 64.5 = 5.5 \]
Rule the median line at its own value
Why: not at the middle of the box.
\[ \text{line at } 66, \quad 66 - 64.5 = 1.5 \text{ from the left end} \]
Draw the lower whisker
Why: from Q1 down to the minimum.
\[ 64.5 - 59 = 5.5 \text{ inches long} \]
Draw the upper whisker
Why: from Q3 up to the maximum.
\[ 77 - 70 = 7 \text{ inches long} \]
Figure (svg): The box plot of the forty heights on a number line from 58 to 78 inches. The box runs from 64.5 to 70, the median line sits at 66, nearer the left end of the box, and the whiskers reach 59 and 77. Labels above give the five numbers. Below, forty dots stack over their heights, with a tall column of nine at 65 and five at 70.
In context: half of the class lies within a 5.5-inch band from 64.5 to 70 inches, the shortest quarter is spread over 5.5 inches below that, and the tallest quarter over 7 inches above it. The median line sits well left of centre because nine students share the height 65.
Check that the drawn lengths add up to the range
Why: whisker, box and whisker cover the data from end to end.
\[ 5.5 + 5.5 + 7 = 18 = 77 - 59 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 95-96 — Example 2.23 and Figure 2.12
Picture it
The book puts one note in bold type: it is important to start a box plot with a scaled number line, otherwise the box plot may not be useful. This picture shows why, using the heights.
Figure (svg): Two drawings of the same five numbers, 59, 64.5, 66, 70 and 77. Top, in red: the five marks spaced evenly with the numbers simply written under them, so the median sits exactly in the middle of the box and both whiskers are as long as each half of the box. Bottom: the same numbers on a scaled number line from 58 to 78 inches, where the median sits left of centre and the upper whisker is longer than the lower.
The top drawing has all five numbers on it, and every one is correct. But it spaces them evenly, so it claims the four parts of the data are equally spread and the median sits in the middle of the box. Both claims are false for this class.
On the scaled line the second quarter is squeezed into 1.5 inches and the top quarter is stretched over 7. That difference is the most interesting fact about these heights, and the unscaled drawing deletes it. A box plot's information is entirely in its positions and widths, so a drawing without a scale has no information left.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 95 — the note on starting with a scaled number line
Trap
A first attempt centres the line in the box, because that looks tidy:
\[ \frac{64.5 + 70}{2} = 67.25 \]
\[ \text{median line at } 67.25 \]
Wrong. 23 heights lie below 67.25 and only 17 above.
The midpoint of the box is the average of two quartiles. It describes the box, not the data, and it puts the line where the eye expects it rather than where half the students are.
The median is a count: half the data on each side. It goes where the data put it, found from places 20 and 21.
\[ \text{median} = \frac{66 + 66}{2} = 66 \]
A line at 66 leaves twenty heights on each side, counting the two 66s as one on each side. A median off centre in its box is information: it says the data inside the box are lopsided.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 95 — the median may sit anywhere between the quartiles
Worked example
Draw the box plot of the forty page counts, whose summary is 136, 237, 322, 395 and 512 pages, and state the interquartile range.
Page counts run into the hundreds, so a scale in steps of 50 pages keeps the ticks readable. Then measure each piece, because the lengths are what you will read off the finished plot.
Choose the scale
Why: it must cover 136 to 512.
\[ \text{axis: } 100 \text{ to } 550, \quad \text{ticks every 50 pages} \]
Box from Q1 to Q3
Why: the interquartile range the book asks for.
\[ IQR = 395 - 237 = 158 \text{ pages} \]
Median line
Why: at its own value inside the box.
\[ 322 - 237 = 85, \qquad 395 - 322 = 73 \]
Lower whisker
Why: from Q1 down to the minimum.
\[ 237 - 136 = 101 \text{ pages} \]
Upper whisker
Why: from Q3 up to the maximum.
\[ 512 - 395 = 117 \text{ pages} \]
Figure (svg): A box plot of the page counts of 40 books on a number line from 100 to 550 pages. The box runs from 237 to 395 with the median line at 322; whiskers reach 136 and 512. Brackets above give the lengths: lower whisker 101, box 158, upper whisker 117. Below the plot, a short orange tick marks each of the forty page counts.
In context: the middle twenty books span 158 pages, from 237 to 395. The median sits a little right of centre in the box, and the two whiskers together, 218 pages, are longer than the box, so the shortest and longest quarters of the shelf are more spread out than the middle half.
Check that the pieces cover the range
Why: three lengths end to end.
\[ 101 + 158 + 117 = 376 = 512 - 136 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Try It 2.23
Error analysis
Here are four drawings of the heights summary, 59, 64.5, 66, 70 and 77, described line by line. Each drawing contains exactly one mistake. Find it in each before you hover over the line to see what went wrong.
Annotate
On: \( \begin{aligned} &(1)\; \text{box drawn from } 59 \text{ to } 77 \\ &(2)\; \text{median line at } 67.25 \\ &(3)\; \text{four parts drawn equally wide} \\ &(4)\; \text{upper whisker drawn to } 78 \end{aligned} \)
Mistakes two and three are the same mistake in different clothes: each draws what you might expect to see rather than what the numbers say. A box plot drawn from expectations always looks symmetric and always hides the one thing worth noticing.
Mistakes one and four come from losing track of what each piece stands for. The box is the middle half; the whiskers reach real data values, never the edge of the paper. Name each piece's number as you draw it and both errors disappear.
Matching
A box plot has five landmarks and each comes from exactly one number in the summary. Match them, then check your matches against the heights plot.
This is the vocabulary for everything that follows, so it pays to be exact. The box has two ends and the whiskers have two ends; only one of the five numbers is drawn inside the box.
Match the pairs
Why: Reading from low to high values along the axis, the landmarks come in the same order as the summary: whisker end, box end, median line, box end, whisker end.
On the heights plot, that is 59, 64.5, 66, 70 and 77. Two of the five match up with the box and two with the whiskers, and only the median is drawn inside. If you ever lose track, read the plot from left to right and say the five names in order.
One caution for later: in a modified plot, Part 4, the whisker ends are no longer the minimum and maximum when an outlier is present. The matching above is for the basic box plot the book draws first.
Worked example
The book describes a box plot where the minimum and Q1 are both 1, the median and Q3 are both 5, and the maximum is 7. Here is a set of eight values with exactly those five numbers. Confirm the summary and say what the picture means.
\[ 1, 1, 1, 5, 5, 5, 5, 7 \]
Ties are the case students find strangest, because pieces of the plot vanish. Work it out by places as usual and the vanishing pieces explain themselves.
Median of eight values
Why: average places 4 and 5.
\[ \text{median} = \frac{5 + 5}{2} = 5 \]
Q1: the lower four, places 1 to 4
Why: average places 2 and 3.
\[ Q_1 = \frac{1 + 1}{2} = 1 = \text{min} \]
Q3: the upper four, places 5 to 8
Why: average places 6 and 7.
\[ Q_3 = \frac{5 + 5}{2} = 5 = \text{median} \]
Figure (svg): A box plot of the eight values 1, 1, 1, 5, 5, 5, 5 and 7 on a number line from 0 to 8. There is no lower whisker: the box starts at 1, which is both the minimum and Q1. The yellow median line lies on the right edge of the box at 5, where Q3 also sits. A single whisker runs from 5 to 7. Below, three dots stack at 1, four at 5 and one at 7.
In context: there is no lower whisker, because the lowest quarter has no spread at all; at least 25 percent of the values equal 1. The median line lies on the right edge of the box, because at least 25 percent of the values equal 5. The top quarter runs from 5 to 7.
Check the book's claim of at least 25 percent at 1
Why: count the ones.
\[ \frac{3}{8} = 37.5\% \ge 25\% \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Figure 2.13 and the paragraph before it
Anomaly
A box plot arrives with its median line lying exactly on the left end of the box. Nothing else is unusual. Decide what this tells you before you read the explanation.
Predict first
The median line coincides with Q1. What must be true of the data?
Correct: At least a quarter of the values equal Q1, so the second quarter has no width
Why: Between Q1 and the median lies about a quarter of the data. If the two cuts coincide, that whole quarter sits on a single value. The quarter is not missing; it is packed into zero width.
The book says this outright: the median can be between the first and third quartiles, or it can be one, or the other, or both. Each coincidence is a quarter of the data stacked on one value, which is common when values are whole numbers or rounded.
Try It 2.22 below has this shape. Ten students report hours on a project, four of them say 3, and both Q1 and the median come out at 3. Wherever you see a line on an edge, think of a pile of identical answers.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 94-95 — the median can coincide with a quartile
Section
Part 3
Concept
Here is the single idea that makes a box plot readable. The four parts, lower whisker, lower half of the box, upper half of the box and upper whisker, each hold about a quarter of the data. That is guaranteed, because the quartiles were chosen by counting.
So the parts cannot differ in how many values they hold. They differ only in how much room those values take up. A narrow part means a quarter of the data packed tightly together; a wide part means a quarter of the data spread thinly.
\[ \text{each part} \approx \tfrac14 \text{ of the data}, \quad \text{width} = \text{how spread out that quarter is} \]
This is what the book means by a good graphical image of the concentration of the data. Concentration is how tightly values crowd, and in a box plot you read it backwards from width: the narrower the part, the higher the concentration.
The reading people get wrong is the opposite one. A long whisker looks like a lot of something, and the eye takes it for a lot of data. It is a lot of distance, holding the same quarter as every other part.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 94 — box plots show the concentration of the data
Worked example
For the forty heights, with summary 59, 64.5, 66, 70 and 77, find the spread of each quarter, the range and the interquartile range.
Each spread is the difference between two neighbouring numbers of the summary. Do them in order from left to right, so that the four answers line up with the four parts of the plot.
(a) Shares
Why: each quarter holds about a quarter of the forty.
\[ \tfrac14 \times 40 = 10 \text{ students per quarter} \]
(b) First quarter
Why: minimum to Q1.
\[ 64.5 - 59 = 5.5 \]
Second quarter
Why: Q1 to the median.
\[ 66 - 64.5 = 1.5 \]
Third quarter
Why: the median to Q3.
\[ 70 - 66 = 4 \]
Fourth quarter
Why: Q3 to the maximum.
\[ 77 - 70 = 7 \]
(c) Range
Why: maximum minus minimum.
\[ 77 - 59 = 18 \]
(d) Interquartile range
Why: Q3 minus Q1.
\[ IQR = 70 - 64.5 = 5.5 \]
In context: about ten students fit in the 1.5 inches from 64.5 to 66, while the tallest ten are spread over 7 inches. The second quarter is the most concentrated and the fourth the least, and the middle half of the class spans 5.5 inches.
Check that the four spreads make up the range
Why: the quarters are consecutive and cover the data end to end.
\[ 5.5 + 1.5 + 4 + 7 = 18 = \text{range} \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Example 2.23(a)-(d)
Picture it
Here are the forty heights again, each dot coloured by which quarter of the places it occupies. Count the dots of each colour before reading on.
Figure (svg): The heights box plot with a bracket above each of its four parts. The brackets are labelled with widths 5.5 inches, 1.5, 4 inches and 7 inches, and each is labelled 10 students. Below, the forty dots are coloured by place: the ten shortest blue, the next ten orange, the next ten purple, the ten tallest green. The orange dots crowd into the narrow strip from 65 to 66; the green dots spread from 70 to 77.
There are ten of each colour. The orange quarter is squeezed into the inch and a half from 64.5 to 66, mostly because nine students are exactly 65 inches. The green quarter is spread over seven inches, with the tallest students well separated from each other.
Now cover the dots and look only at the box plot. The narrow piece and the wide piece are still there, and they still mean crowded and spread out. The box plot keeps the concentration and throws away the individual dots, which is exactly the trade it was designed to make.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Example 2.23(b)
Worked example
Compare the interval from 59 to 65 inches with the interval from 66 to 70. Which holds more of the data? Then describe the middle half.
Widths cannot answer this, since the first interval is 6 inches wide and the second is 4. What decides it is where the quartiles fall inside each interval. Locate each interval against the summary first, then count.
Locate 66 to 70 on the summary
Why: it runs from the median to Q3.
\[ 66 \text{ to } 70 = \text{median to } Q_3 \;\Rightarrow\; \text{about one quarter} \]
Locate 59 to 65
Why: it covers the whole first quarter and passes Q1 = 64.5.
\[ 59 \text{ to } 65 \supset 59 \text{ to } 64.5 \;\Rightarrow\; \text{more than one quarter} \]
Count them exactly
Why: from the sorted list, both ends included.
\[ 59 \text{ to } 65: \; 19 \text{ of } 40 = 47.5\% \]
The other interval
Why: the two 66s, two 67s, two 68s, one 69 and five 70s.
\[ 66 \text{ to } 70: \; 12 \text{ of } 40 = 30\% \]
(f) The middle half
Why: the box.
\[ IQR = 5.5 \text{ inches} \]
In context: the interval from 59 to 65 holds more students, nearly half the class, even though it is only 2 inches wider. The middle half of the class spans 5.5 inches. The book says 66 through 70 holds 25 percent; counted with both ends included it holds 30 percent, because the ties at 66 and 70 belong to it.
Check the conclusion survives the correction
Why: 19 against 12.
\[ 47.5\% > 30\% \quad \text{the book's comparison stands} \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Example 2.23(e)-(f)
Concept
The book is careful to say each quarter holds approximately 25 percent of the data. The word approximately matters, and the heights show why. By places, each quarter holds exactly ten students. By values, students who share a height cannot be split across a cut.
\[ \text{below } 64.5: 10, \quad 64.5 \text{ to under } 66: 9, \quad 66 \text{ to under } 70: 7, \quad 70 \text{ and up: } 14 \]
The median is 66, and two students are exactly 66 inches tall. Counting by places, one of them belongs to the lower half and one to the upper half. Counting by values, both land on the same side of whichever rule you choose. The same happens at Q3, where five students are exactly 70.
\[ 10 + 9 + 7 + 14 = 40, \qquad \text{but by places } 10 + 10 + 10 + 10 = 40 \]
So when a question asks what share lies between two cuts, the honest answer is about a quarter, and an exact count needs the data and a decision about the tied values. The book's own Example 2.24(c) trips over exactly this, and you will see the correction in Part 5.
None of this changes how you read the plot. Each part is still roughly a quarter, a narrow part is still crowded, and a part with no width at all, like the second quarter in Figure 2.13, is a quarter of the data piled on one value.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 96 — Example 2.23(a), each quarter has approximately 25 percent
Ranking
The four quarters of the heights span 5.5, 1.5, 4 and 7 inches, and each holds ten students. Put them in order from the most concentrated to the least.
Most concentrated means most tightly packed: the most students per inch. Since every quarter has the same ten students, decide what the width alone tells you before you drag anything.
Put in order
Drag the quarters into order, most crowded first.
Why: Same count in each, so the narrowest is the most crowded: 1.5 inches, then 4, then 5.5, then 7. The second quarter packs ten students into an inch and a half.
If you put the widest first, you ranked by room instead of by crowding. The question asks how tightly the students are packed, and with ten in every quarter, tight packing means a short width.
A quick way to say it in numbers: ten students in 1.5 inches is about 6.7 students per inch; ten in 7 inches is about 1.4 per inch. Dividing the count by the width turns a box plot's widths into densities, and the ranking falls out at once.
Trap
Looking at the Example 2.25 box plot, a first reading says:
\[ \text{upper whisker: } 790 - 490 = 300 \]
\[ \text{lower whisker: } 15 - 10 = 5 \]
Wrong: most of the data must be up at the top.
This is the most common misreading of a box plot, because a long shape looks like a big quantity. The eye reads length as amount.
Each whisker covers about a quarter of the values. Count them from the sorted list instead of measuring.
\[ \text{at or above } Q_3 = 490: \; 4 \text{ of } 15 \]
\[ \text{at or below } Q_1 = 15: \; 4 \text{ of } 15 \]
The same four values in each, so the long whisker is a quarter of the data spread thinly over 300 units. Length is room, not count.
Change 790 to 7,900 and the whisker grows tenfold while still holding the same four values. One value drew that length; four values share it.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 98 — Example 2.25
Worked example
Graph a box-and-whisker plot for these fifteen values, already sorted.
\[ 10, 10, 10, 15, 35, 75, 90, 95, 100, 175, 420, 490, 515, 515, 790 \]
Fifteen is odd, so the median is a single value and it is left out of both halves. That makes every cut a real data value, which is the tidy case.
Median place
Why: fifteen values.
\[ \frac{15 + 1}{2} = 8 \;\Rightarrow\; \text{median} = 95 \]
Q1: the lower seven, places 1 to 7
Why: its middle is the 4th place.
\[ Q_1 = 15 \]
Q3: the upper seven, places 9 to 15
Why: its middle is the 12th place.
\[ Q_3 = 490 \]
Measure the two whiskers
Why: box ends to the extremes.
\[ 15 - 10 = 5, \qquad 790 - 490 = 300 \]
Measure the two halves of the box
Why: the median sits near the left end.
\[ 95 - 15 = 80, \qquad 490 - 95 = 395 \]
Figure (svg): A box plot of the fifteen values of Example 2.25 on a number line from 0 to 800. The box runs from 15 to 490 with the median at 95, close to the left end of the box. The lower whisker, from 10 to 15, is almost invisible; the upper whisker runs all the way from 490 to 790. Labels say each whisker covers about a quarter of the values: four values below Q1 counting the minimum, four from Q3 up. Fifteen dots below, crowded at the left.
Compare the two halves of the box as well as the whiskers. The lower half spans 80 units and the upper half 395, so the same pattern repeats inside the box: crowded at the low end, spread out at the high end.
In context: the lower half of the values is packed between 10 and 95, while the upper half stretches from 95 to 790. Everything about the plot leans the same way, long on the right, and Section 2.6 will call that shape skewed to the right.
Check against the book
Why: its printed five numbers on p. 98.
\[ 10, \; 15, \; 95, \; 490, \; 790 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 98 — Example 2.25 and Figure 2.15
Concept
A box plot also tells you whether the data are balanced around the middle. Compare the two halves of the box, and then the two whiskers. If both comparisons lean the same way, the data are stretched out on that side.
Example 2.25 leans right everywhere. The lower half of its box is 80 units wide and the upper half 395; the lower whisker is 5 units long and the upper whisker 300. The small values are crowded together and the large ones are spread far apart.
\[ \text{box halves: } 95 - 15 = 80 \; < \; 490 - 95 = 395 \]
\[ \text{whiskers: } 15 - 10 = 5 \; < \; 790 - 490 = 300 \]
The day class of Example 2.24 leans the other way. Its median sits close to Q3, and its lower whisker is the longer one: the higher scores are packed together and a few low scores trail far below.
\[ \text{box halves: } 74.5 - 56 = 18.5 \; > \; 82.5 - 74.5 = 8 \]
\[ \text{whiskers: } 56 - 32 = 24 \; > \; 99 - 82.5 = 16.5 \]
When the halves of the box and the whiskers are close to equal, as for Try It 2.23's books, the data are roughly balanced. Section 2.6 gives these shapes names, skewed right, skewed left and symmetric; for now, describe them in plain words, with the numbers that show it.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 97-98 — Examples 2.24 and 2.25
Discrimination
Every feature of a box plot reports something, and it is never a count. Sort each feature by what it actually tells you about the data.
Ask of each feature: is it a single width, a position on the axis, or a comparison of two widths? A single width is about spread; a position is about centre; a comparison is about shape.
Sort into buckets
Sort each feature of a box plot by what it reports.
Notice that nothing in this list reports how many values there are. A box plot of 15 values and one of 15,000 look alike if their five numbers agree. If the sample size matters, and it usually does, it has to be written beside the plot.
The comparisons in the third bucket are the start of Section 2.6's skewness. For now, read them in words: stretched further to the right, or packed against the low end of the box.
Section
Part 4
Concept
A basic box plot always stretches its whiskers to the very ends of the data, so one unusual value can drag a whisker across the whole picture. To decide whether a value is unusual, the book measures it against the width of the box.
\[ \text{lower fence} = Q_1 - 1.5\,(IQR), \qquad \text{upper fence} = Q_3 + 1.5\,(IQR) \]
Any value below the lower fence or above the upper fence is a potential outlier. The rule uses the IQR as its ruler because the IQR is resistant: the suspected value cannot stretch the ruler that is used to judge it. A ruler made from the range would grow with the very value it was measuring.
Figure (svg): The exercise data's box from Q1 = 20 to Q3 = 60, labelled IQR = 40, with the median at 40. Orange arrows reach 60 minutes beyond each quartile, labelled 1.5 times 40 equals 60, to dashed red fences at minus 40 and at 120. Fifteen grey dots sit along the axis from 0 to 120, and one red dot at 300 lies beyond the upper fence.
For the exercise data the box is 40 minutes wide, so each fence sits 60 minutes beyond its quartile: at minus 40 and at 120. No one can exercise a negative number of minutes, so nothing can be a low outlier here. The value 300 is far beyond 120.
potential outlier — A value more than 1.5 interquartile ranges below Q1 or above Q3. It is a flag for a closer look, not proof that the value is wrong.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 94 — Example 2.22, the outlier test the book uses
Notation
Every symbol in the fence formula is something you have already met. Hover over each piece; the one people misread is the 1.5.
Annotate
On: \( Q_3 + 1.5\,(IQR) = 60 + 1.5(40) = 120 \)
Say it in words as you write it: the upper fence is Q3 plus one and a half box-widths. Saying box-widths reminds you that the multiplier applies to the IQR, not to Q3, which is the commonest slip when the formula is typed into a calculator as 1.5 times 60 plus 40.
The lower fence mirrors it, Q1 minus one and a half box-widths. Always compute both. For measurements that cannot be negative the lower fence is often below zero, which just means there can be no low outliers.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 94 — Q3 + 1.5(IQR) in Example 2.22
Intuition
Suppose you tried to judge outliers with the range instead, flagging anything more than, say, a quarter of the range beyond Q3. For the exercise data the range is 300, so the cutoff would sit at 60 plus 75, which is 135, and 300 would still be flagged.
\[ \text{range-based: } 60 + \tfrac14(300 - 0) = 135 \]
Now make the unusual value even more extreme, 600 minutes instead of 300. The range doubles, the cutoff jumps to 210, and it keeps chasing the value it is supposed to catch. The more extreme the value, the more lenient the rule becomes.
\[ \text{range-based, with 600: } \; 60 + \tfrac14(600) = 210 \]
\[ \text{IQR-based, either way: } \; 60 + 1.5(40) = 120 \]
The IQR does not have that flaw. The value 300 or 600 sits at place 15, outside the middle half, so the box and its width stay exactly the same, and the fence stays at 120. The ruler is made from the part of the data the suspect value cannot reach.
That is the same resistance you met in Part 1, now put to work. Whenever you test whether a value is unusual, measure it against something it could not have distorted.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 94 — Example 2.22, Q3 + 1.5(IQR) = 120
Worked example
A principal surveys 15 anonymous students about how many minutes a day they exercise, to decide whether to buy new fitness equipment. Find the five-number summary.
\[ 0, 40, 60, 30, 60, 10, 45, 30, 300, 90, 30, 120, 60, 0, 20 \]
The values arrive in the order the students answered, so sorting comes first. Fifteen values means the cuts fall at places 4, 8 and 12, all whole places.
Sort the fifteen values
Why: smallest to largest.
\[ 0, 0, 10, 20, 30, 30, 30, 40, 45, 60, 60, 60, 90, 120, 300 \]
Median at place 8
Why: fifteen values; seven on each side.
\[ \text{median} = 40 \]
Q1 at place 4
Why: the middle of places 1 to 7.
\[ Q_1 = 20 \]
Q3 at place 12
Why: the middle of places 9 to 15.
\[ Q_3 = 60 \]
Extremes and IQR
Why: first and last; Q3 minus Q1.
\[ \text{min} = 0, \quad \text{max} = 300, \quad IQR = 60 - 20 = 40 \]
Several students gave identical answers, three 30s and three 60s, which is typical of rounded survey responses. Q3 lands on one of the three 60s, so at least a quarter of the students reported 60 minutes or more.
In context: 75 percent of the students exercise 60 minutes a day or less, and the middle half exercise between 20 and 60 minutes. The principal judged that reasonable and decided the equipment was justified.
Check against the book
Why: its printed values on p. 94.
\[ 0, \; 20, \; 40, \; 60, \; 300 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, pp. 93-94 — Example 2.22
Worked example
Is 300 minutes a potential outlier? Draw the modified box plot, and say what happens to the summary if the value is removed.
A modified box plot changes only the whiskers. Find the fences, flag what lies beyond them, and end each whisker at the most extreme value that is still inside.
Upper fence
Why: Q3 plus one and a half IQRs.
\[ 60 + 1.5(40) = 60 + 60 = 120 \]
Lower fence
Why: Q1 minus one and a half IQRs.
\[ 20 - 1.5(40) = 20 - 60 = -40 \]
Test the extremes
Why: 300 is beyond the upper fence; 0 is inside the lower.
\[ 300 > 120 \;\Rightarrow\; \text{potential outlier} \]
End the upper whisker
Why: the largest value not beyond the fence.
\[ \text{largest value} \le 120 \text{ is } 120 \]
Figure (svg): Two box plots of the exercise minutes on one axis from 0 to 320. Top, labelled basic: the box from 20 to 60 with median 40, and the upper whisker stretching all the way to 300. Bottom, labelled modified: the same box, the upper whisker stopping at 120, and a hollow red circle at 300. A dashed red line marks the fence at 120. Fifteen dots sit along the bottom.
Remove 300 and recompute, as the book does
Why: fourteen values; median between places 7 and 8.
\[ \text{median} = \frac{30 + 40}{2} = 35, \quad Q_1 = 20, \quad Q_3 = 60, \quad \text{max} = 120 \]
In context: the conclusion survives, three quarters still exercise 60 minutes or less, but 15 students is a small sample and the principal should survey more. The book lists the new minimum, quartiles and maximum but leaves out the median, which falls from 40 to 35.
Check the modified plot loses nothing
Why: the box and median are identical in both drawings; only the whisker moved.
\[ Q_1 = 20, \; M = 40, \; Q_3 = 60 \text{ in both plots} \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 94 — Example 2.22; §2.4, p. 95, the note on dots marking outliers
Worked example
Follow the steps for the fifteen values below, and decide whether any value should be drawn as a separate point.
\[ 0, 5, 5, 15, 30, 30, 45, 50, 50, 60, 75, 110, 140, 240, 330 \]
The book asks only for the basic plot. Add the fence test as well, because with a maximum of 330 and a Q3 near 100 the top whisker looks suspicious, and the fence settles it.
Places 8, 4 and 12
Why: fifteen values, already sorted.
\[ \text{median} = 50, \quad Q_1 = 15, \quad Q_3 = 110 \]
Interquartile range
Why: Q3 minus Q1.
\[ IQR = 110 - 15 = 95 \]
Upper fence
Why: one and a half IQRs above Q3.
\[ 110 + 1.5(95) = 110 + 142.5 = 252.5 \]
Lower fence
Why: one and a half IQRs below Q1.
\[ 15 - 142.5 = -127.5 \]
Flag and end the whisker
Why: 330 is beyond 252.5; 240 is the largest value inside.
\[ \text{outlier: } 330, \qquad \text{upper whisker ends at } 240 \]
Figure (svg): A modified box plot of the fifteen values of Try It 2.25 on a number line from 0 to 350. The box runs from 15 to 110 with the median at 50. The lower whisker reaches 0. The upper whisker stops at 240, short of a dashed red fence at 252.5, and a hollow red circle marks 330 beyond it. Fifteen dots below.
In context: the middle half of the values lies between 15 and 110, and 330 stands well apart from the rest. The basic plot would run its whisker to 330 and hide that; the modified plot shows it as a point on its own.
Check the whisker end is a data value
Why: 240 is in the list, and it is inside the fence.
\[ 240 \le 252.5 < 330 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 98 — Try It 2.25
Trap
Having found the fence for Try It 2.25, a first attempt draws:
\[ \text{fence} = 252.5 \]
\[ \text{upper whisker from } 110 \text{ to } 252.5 \]
Wrong. No value in the data is 252.5.
The mistake is natural, because the fence is the last number computed and it feels like the answer. But the fence is a test, not a landmark.
The fence is a threshold for testing values. A whisker must end at a real observation: the most extreme value that is not beyond the fence.
\[ \text{whisker ends at } 240 \]
A whisker drawn to 252.5 claims the data reach a value nobody recorded. In Example 2.22 the whisker end and the fence were both 120, which is a coincidence of that data set, not the rule.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 95 — the note on box plots with dots marking outliers
Concept
The book never says that 300 minutes is wrong. It says 300 appears to be a potential outlier, and then it asks the principal to be careful. A value beyond a fence has three possible stories, and the fence cannot tell you which one is true.
The book's own move in Example 2.22 is to recompute without the 300 and see whether the conclusion changes. It does not: three quarters of students still exercise 60 minutes or less. Checking whether a conclusion depends on one value is honest; quietly deleting the value is not.
The modified box plot is the display version of that honesty. It keeps the outlier in the data and in every calculation, so the quartiles are unchanged, and it draws the value as a separate point so the reader can see it and ask the question for themselves.
\[ \text{outlier flagged} \;\Rightarrow\; \text{still counted in } n, Q_1, M, Q_3 \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 94 — Example 2.22, the principal needs to be careful
Sorting
A data set of waiting times in minutes has Q1 = 10 and Q3 = 30. Before sorting, work out the fences, then decide where each value would be drawn in a modified box plot.
\[ IQR = 20, \quad \text{fences: } 10 - 30 = -20 \text{ and } 30 + 30 = 60 \]
Sort into buckets
Sort each value by how a modified box plot draws it.
Notice how close 58 and 62 are. The fence is a sharp line, so two nearly equal values can land on opposite sides of it. That is one reason the book says potential outlier: the label is a prompt to look, not a verdict about the value.
Notice also that 0 is fine. The lower fence, minus 20, is below anything a waiting time can be, so this data set cannot have a low outlier no matter what is recorded.
Worked example
An instructor wants the typical time on a final project to be about 3 to 4 hours. A random sample of 10 students reports these hours. Find the five numbers, test for outliers, and advise the instructor.
\[ 2, 3, 5, 3, 4, 4, 3, 11, 3, 2 \]
Ten is even, so each half holds five values and the quartiles land on actual values. Several students gave the same answer, so expect some of the five numbers to coincide.
Sort
Why: smallest to largest.
\[ 2, 2, 3, 3, 3, 3, 4, 4, 5, 11 \]
Median: average places 5 and 6
Why: both are 3.
\[ \text{median} = \frac{3 + 3}{2} = 3 \]
Q1 and Q3: the 3rd and 8th places
Why: the middles of the two halves of five.
\[ Q_1 = 3, \qquad Q_3 = 4 \]
IQR and upper fence
Why: one and a half box-widths above Q3.
\[ IQR = 1, \quad 4 + 1.5(1) = 5.5 \]
Flag the outlier and end the whisker
Why: 11 is beyond 5.5; the largest value inside is 5.
\[ \text{outlier } 11, \quad \text{whisker ends at } 5 \]
Figure (svg): A modified box plot of the ten project times in hours on a number line from 0 to 12. The box runs from 3 to 4, with the median line on its left edge at 3. Whiskers reach 2 and 5. A dashed red fence sits at 5.5, and a hollow red circle marks 11. Dots below: two at 2, four at 3, two at 4, one at 5 and one at 11.
In context: the median student spent 3 hours and the middle half spent 3 to 4 hours, exactly the instructor's target, so the project can stay as it is. The one student who reported 11 hours is worth a conversation, not a redesign.
Check the summary in order
Why: min, Q1, median, Q3, max.
\[ 2 \le 3 = 3 \le 4 \le 11 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.3, p. 94 — Try It 2.22
Section
Part 5
Concept
A box plot is short and flat, so several fit one above another against a single number line. That is where box plots earn their keep. Two histograms on one set of axes cover each other up; box plots simply stack.
The book's Example 2.24 asks for exactly this: create a box plot for each set of data, and use one number line for both. The shared axis is not a style choice. Every comparison you make, a median further right, a box twice as wide, is a comparison of positions and lengths, and that only means something if both plots use the same scale.
Read the plots in that order and write one sentence per line, naming both groups and the numbers. A comparison that says only that one group is bigger has not said which feature is bigger, and that is the whole point of the picture.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Example 2.24(d), one number line for both
Step zero
Test scores for a day statistics class (20 students) and an evening class (22 students) are listed below. The example asks for both five-number summaries, the share of data between each pair of cuts, parallel box plots, and which class has the wider middle 50 percent.
\[ \text{day: } 99, 56, 78, 55.5, 32, 90, 80, 81, 56, 59, \]
\[ 45, 77, 84.5, 84, 70, 72, 68, 32, 79, 90 \]
\[ \text{evening: } 98, 78, 68, 83, 81, 89, 88, 76, 65, 45, 98, \]
\[ 90, 80, 84.5, 85, 79, 78, 98, 90, 79, 81, 25.5 \]
Discussion prompt
Before any arithmetic: which places will hold the median, Q1 and Q3 for each class, and which of the questions can be answered from the box plots alone?
Planning the places first is worth the minute it takes. The two classes have different sizes, so their quartiles sit at different places, and an even half versus an odd half changes whether you average or read one value.
Planning also tells you how much work each part is. Parts (a), (b) and (d) need only the sorted lists and five numbers each. Part (c) needs real counting, and part (e) needs only two subtractions once the summaries exist. Do the sorting once, carefully, and everything else follows from it.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Example 2.24
Worked example
Find the smallest and largest values, the median, and the first and third quartiles for the day class.
Twenty values, so every cut averages two neighbours. Sort carefully: the two 32s and the two 56s are easy to drop.
Sort the twenty scores
Why: smallest to largest.
\[ 32, 32, 45, 55.5, 56, 56, 59, 68, 70, 72, 77, 78, 79, 80, 81, 84, 84.5, 90, 90, 99 \]
Median: places 10 and 11
Why: 72 and 77.
\[ \text{median} = \frac{72 + 77}{2} = \frac{149}{2} = 74.5 \]
Q1: places 5 and 6
Why: 56 and 56.
\[ Q_1 = \frac{56 + 56}{2} = 56 \]
Q3: places 15 and 16
Why: 81 and 84.
\[ Q_3 = \frac{81 + 84}{2} = \frac{165}{2} = 82.5 \]
Extremes
Why: first and last.
\[ \text{min} = 32, \qquad \text{max} = 99 \]
Notice that Q1 came out as a real score, 56, because two students tied at places 5 and 6, while the median and Q3 are scores nobody earned. Both are normal. What matters is that ten scores lie on each side of 74.5 and four lie strictly below the cut at 56.
In context: the middle half of the day class scored between 56 and 82.5, a spread of 26.5 points, and the median score was 74.5. The lowest quarter reaches all the way down to 32.
Check against the book
Why: its printed summary.
\[ 32, \; 56, \; 74.5, \; 82.5, \; 99 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Example 2.24(a)
Worked example
Find the same five numbers for the evening class, which has 22 students.
Twenty-two values: the median averages places 11 and 12, and each half has eleven values, so Q1 and Q3 are single places, the 6th and the 17th.
Sort the twenty-two scores
Why: smallest to largest.
\[ 25.5, 45, 65, 68, 76, 78, 78, 79, 79, 80, 81, \]
\[ 81, 83, 84.5, 85, 88, 89, 90, 90, 98, 98, 98 \]
Median: places 11 and 12
Why: both 81.
\[ \text{median} = \frac{81 + 81}{2} = 81 \]
Q1: place 6, the middle of the lower eleven
Why: count from the left.
\[ Q_1 = 78 \]
Q3: place 17, the middle of the upper eleven
Why: places 12 to 22; the 6th of them.
\[ Q_3 = 89 \]
Extremes
Why: first and last.
\[ \text{min} = 25.5, \qquad \text{max} = 98 \]
The odd-sized halves are the step to watch. Eleven values in each half means the middle of each half is a single place, the 6th of eleven, so no averaging is needed for Q1 or Q3, only for the median. Counting the upper half from place 12 gives its 6th value at place 17.
In context: the middle half of the evening class scored between 78 and 89, a spread of only 11 points, around a median of 81. Two scores, 25.5 and 45, sit far below everyone else.
Check against the book
Why: its printed summary.
\[ 25.5, \; 78, \; 81, \; 89, \; 98 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Example 2.24(b)
Worked example
For each class, what percentage of the data lies between each pair of neighbouring cuts, and between Q1 and the largest value?
By places the answer is always about 25 percent per part. Counting actual values is where ties bite, and the book's own count stumbles on them, so count each value exactly once, deciding where tied values go.
Day class: below Q1 = 56
Why: 32, 32, 45, 55.5.
\[ 4 \text{ of } 20 = 20\% \]
From 56 up to the median 74.5
Why: 56, 56, 59, 68, 70, 72.
\[ 6 \text{ of } 20 = 30\% \]
The two upper parts
Why: 77 to 81, then 84 to 99.
\[ 5 \text{ of } 20 = 25\%, \qquad 5 \text{ of } 20 = 25\% \]
From Q1 to the largest
Why: every value of 56 or more.
\[ 16 \text{ of } 20 = 80\% \]
Evening class, the same way
Why: below 78; 78 to 80; 81 to 88; 89 and up.
\[ 5, \; 5, \; 6, \; 6 \text{ of } 22 = 23\%, \; 23\%, \; 27\%, \; 27\% \]
The book counts six values from 32 to 56 and six from 56 to 74.5, so the two 56s are counted twice and its four shares add to 110 percent. It then calls 16 of 20 values 75 percent; 16 of 20 is 80 percent. Each part holds about a quarter by places; exact counts drift because two students tied at Q1.
Check that every value is counted once
Why: the four parts must total n.
\[ 4 + 6 + 5 + 5 = 20, \qquad 5 + 5 + 6 + 6 = 22 \quad \checkmark \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Example 2.24(c); the evening answer is left blank in the book
Worked example
Draw both box plots on one number line, and decide which class has the wider spread for its middle 50 percent. What does that mean?
Both classes' scores run from 25.5 to 99, so a single axis from 20 to 100 holds everything. Draw the two plots one above the other, then compare the boxes.
Shared axis
Why: it must cover both classes.
\[ \text{axis: } 20 \text{ to } 100, \quad \text{ticks every 10 points} \]
Day class IQR
Why: the width of its box.
\[ 82.5 - 56 = 26.5 \]
Evening class IQR
Why: the width of its box.
\[ 89 - 78 = 11 \]
Compare
Why: divide the two widths.
\[ \frac{26.5}{11} \approx 2.4 \]
Figure (svg): Parallel box plots of test scores on one number line from 20 to 100. Day class, above: box from 56 to 82.5, median 74.5, whiskers from 32 to 99, with twenty orange dots below it. Evening class, below: box from 78 to 89, median 81, whiskers from 25.5 to 98, with twenty-two orange dots below it, two of them far to the left at 25.5 and 45.
Look at the figure before reading the numbers. The day box is visibly more than twice as wide, and the evening box sits to the right of the day median. A shared axis lets your eye do the comparison that the subtraction confirms.
In context: the day class's middle half is spread over 26.5 points, about two and a half times the evening class's 11, so the day class's typical scores vary much more. The evening class also sits higher, with a median of 81 against 74.5.
Check the evening outliers with the fences
Why: each class is judged against its own box.
\[ 78 - 1.5(11) = 61.5 \;\Rightarrow\; 25.5 \text{ and } 45 \text{ are potential outliers} \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 97-98 — Example 2.24(d)-(e) and Figure 2.14
Concept
A picture is not yet a conclusion. Once two box plots share an axis, turn them into sentences, one feature at a time, each naming both groups, the numbers and the units. Here is that write-up for the two statistics classes.
Centre. The evening class's median, 81, is 6.5 points higher than the day class's 74.5, so a typical evening student scored higher.
\[ 81 - 74.5 = 6.5 \text{ points} \]
Spread of the middle half. The day class's IQR, 26.5 points, is about two and a half times the evening class's 11, so day-class scores in the middle are far less predictable.
Reach. Both classes have a few very low scores, but in the evening class the two lowest, 25.5 and 45, lie beyond its lower fence of 61.5; in the day class nothing lies beyond a fence.
Overlap. The day class's box, 56 to 82.5, contains the evening class's lower quartile and median, so many students in the two classes scored alike despite the different centres.
Each sentence answers a different question, and none of them alone describes the comparison. Notice too what the sentences do not claim: nothing about why the classes differ. Two classes on one test is an observational comparison, and Chapter 1's warnings about cause apply in full.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 97-98 — Example 2.24(e)
Comparison
Fill in the blanks to finish the comparison table for Example 2.24. Every entry is a difference of two summary numbers you have already found.
Comparison matrix
| measure | day class | evening class | which is larger |
|---|---|---|---|
| median | 74.5 | 81 | evening |
| IQR | 26.5 | 11 | day |
| range | 67 | 72.5 | evening |
| lower fence | 16.25 | 61.5 | evening |
| potential outliers | none | 2 | evening |
Look at the IQR row against the range row. By range, the evening class is the more spread out, 72.5 against 67, entirely because of one score of 25.5. By IQR the day class is more than twice as spread. The two measures disagree because they ask different questions, and the IQR is the one that describes the typical student.
The fences row explains a puzzle: the day class's 32 is not flagged but the evening class's 45 is. Each class is judged against its own box, and the evening class's narrow box puts its lower fence up at 61.5. Unusual always means unusual for this group.
Worked example
Heights in inches for the 20 boys and the 20 girls in a class of 40 are below. Find each five-number summary, draw them on one number line, and say which has the wider middle 50 percent.
\[ \text{boys: } 66, 66, 67, 67, 68, 68, 68, 68, 68, 69, 69, 69, 70, 71, 72, 72, 72, 73, 73, 74 \]
\[ \text{girls: } 61, 61, 62, 62, 63, 63, 63, 65, 65, 65, 66, 66, 66, 67, 68, 68, 68, 69, 69, 69 \]
Both groups have twenty values, so the cuts sit at the same places for both: Q1 averages places 5 and 6, the median places 10 and 11, Q3 places 15 and 16. Only the heights at those places differ.
Boys
Why: places 5 and 6, 10 and 11, 15 and 16.
\[ Q_1 = 68, \quad M = 69, \quad Q_3 = 72, \quad \text{min } 66, \; \text{max } 74 \]
Girls: the median
Why: places 10 and 11 hold 65 and 66.
\[ M = \frac{65 + 66}{2} = 65.5 \]
Girls: the other four
Why: places 5 and 6 both 63; 15 and 16 both 68.
\[ Q_1 = 63, \quad Q_3 = 68, \quad \text{min } 61, \; \text{max } 69 \]
Compare the boxes
Why: the two IQRs.
\[ IQR_{\text{boys}} = 72 - 68 = 4, \qquad IQR_{\text{girls}} = 68 - 63 = 5 \]
Figure (svg): Parallel box plots of heights in inches on one number line from 60 to 76. Boys, above: box from 68 to 72, median 69, whiskers from 66 to 74. Girls, below: box from 63 to 68, median 65.5, whiskers from 61 to 69. A faint band from 66 to 69 marks heights reached by both groups. Dots under each plot.
In context: the girls' middle half is spread over 5 inches and the boys' over 4, so the girls' heights are a little more variable in the middle. The boys' median is 3.5 inches higher. The boxes meet only at 68, yet the whiskers show both groups reaching 66 to 69 inches, so the groups overlap more than the boxes suggest.
Check that the ranges really are equal
Why: the two groups differ inside, not at the ends.
\[ 74 - 66 = 8 = 69 - 61 \]
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 98 — Try It 2.24
Trap
Two box plots arrive on separate axes, each stretched to fill its own width. A reader compares them:
\[ \text{books: box looks huge} \]
\[ \text{heights: box looks huge} \]
Wrong: the two are equally spread.
Software does this by default: each plot gets its own axis, scaled to fill the space. Two pictures that look the same can describe completely different spreads.
On its own axis every plot fills the page, so the drawn lengths say nothing about the data. Compare the numbers, in their units, instead.
\[ IQR = 158 \text{ pages} \]
\[ IQR = 5.5 \text{ inches} \]
Pages and inches cannot be compared at all. Groups measured in the same unit can be, but only after both are drawn on one shared scale.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, p. 97 — Example 2.24(d), one number line
Counterexample
A box plot keeps five numbers and throws away everything else. Here are two sets of eleven values that produce exactly the same box plot: minimum 0, Q1 20, median 50, Q3 80 and maximum 100.
\[ A: \; 0, 10, 20, 22, 24, 50, 76, 78, 80, 90, 100 \]
\[ B: \; 0, 5, 20, 48, 49, 50, 51, 52, 80, 95, 100 \]
Figure (svg): Two dot plots of eleven values each, over one shared box plot. Set A, top: two clusters, one from 0 to 24 and one from 76 to 100, with only the value 50 between them. Set B, middle: values packed from 48 to 52 around 50, plus a few spread out to the ends. Bottom: the single box plot both sets produce, box from 20 to 80 with median 50 and whiskers to 0 and 100.
Check the claim yourself: in both sets the 6th value is 50, the 3rd is 20 and the 9th is 80, and both run from 0 to 100. Every one of the five numbers agrees, so the box plots must be identical, line for line.
Discussion prompt
Describe how the two data sets differ, and name a display from Sections 2.1 or 2.2 that would show the difference.
So a box plot cannot show clusters, gaps, or the number of observations. That is why it is often drawn beside a histogram or with its sample size written next to it. What it does better than anything else is line up many groups on one axis, and that is worth the detail it gives up.
Real world
A hospital publishes parallel box plots of emergency waiting times for three departments, all on one axis in minutes. A manager reads them and says department A is the fastest, because its median is the lowest.
Read the three summaries below as box plots in your head: where is each median, how wide is each box, and how far does each upper whisker reach? Then judge the manager's claim.
Discussion prompt
What does each plot actually say about patients, and is the manager's reading of department A defensible?
The manager read one feature, the centre, and ignored the rest of the picture. A box plot is built to show centre, spread and reach at once, and a responsible reading names all three, in minutes, before it ranks anything.
Section
Part 6
Pattern
Every box plot problem in this section followed the same five moves. The first two are counting and the next two are measuring; the last turns the picture into sentences.
Use the diagram as a checklist and say which kind of step you are on. In steps one and two you are counting places in a list; in steps three and four you are measuring distances on the scale; in step five you are writing sentences with units.
Figure (svg): A flow diagram of five boxes joined by arrows. Sort the data and count n. Find the median, then Q1 and Q3 as the medians of the halves. Compute the IQR and the fences. Draw a scaled axis, the box, the median line and the whiskers. Read and compare: widths as concentration, medians as centre.
\[ IQR = Q_3 - Q_1, \qquad \text{fences: } Q_1 - 1.5\,(IQR), \quad Q_3 + 1.5\,(IQR) \]
Most wrong answers come from mixing up counting and measuring: a quartile found by dividing the scale into four, a median drawn in the middle of the box, a long whisker read as lots of data. Each has a one-line check: count how many values lie on each side.
OpenStax Introductory Business Statistics 2e, §2.1 Display Data §2.1 Display Data — the same construction, in business contexts
Check
Find the three cuts by places before looking at the options.
Fifteen values means the median is the 8th value and each half holds seven. Find the middle place of each half, then compare with the options. Two of the wrong options come from common counting slips, and one from measuring the scale.
Check your understanding
The fifteen values 0, 5, 5, 15, 30, 30, 45, 50, 50, 60, 75, 110, 140, 240 and 330 are already sorted. What is their five-number summary?
Answer: A
Why: The median is the 8th value, 50. The lower half is places 1 to 7, whose middle is the 4th value, 15. The upper half is places 9 to 15, whose middle is the 12th value, 110.
In context: the middle half of these values lies between 15 and 110. The quick check is to count: three values lie below 15 and three above 110, with the quartile values themselves at places 4 and 12, a quarter of the data at each end.
With these five numbers the fences are minus 127.5 and 252.5, so 330 would be drawn as a separate point in a modified plot. A summary is only the first step.
Check
Count, do not measure.
A café's box plot shows the waiting times of 60 orders. Decide what share of the orders the third quartile cuts off before you look at any option, then turn the share into a number of orders.
Check your understanding
A café plots the waiting times of 60 orders. The plot has Q3 at 9 minutes and an upper whisker reaching 21 minutes, the longest part of the plot. About how many orders waited 9 minutes or more?
Answer: A
Why: Q3 cuts off the top quarter of the orders, whatever the length of the whisker. A quarter of 60 is 15, so about 15 orders waited 9 minutes or more.
In context: about 15 of the 60 orders waited at least 9 minutes, and those slowest orders were spread out as far as 21 minutes. The long whisker says the slow orders vary a lot, not that there are many of them.
Notice the answer used the sample size, 60, which the plot itself does not show. That is a reminder of what a box plot leaves out.
The same reasoning gives about 15 orders below Q1 and about 30 on each side of the median, whatever the minutes on the axis say.
Check
Find the fence first, then look at the data values near it.
The fence is a threshold computed from the box. The whisker has to end at a real data value, so after computing the fence, look for the largest value that does not pass it.
Check your understanding
Commute times in minutes have Q1 = 12 and Q3 = 20. The largest four values are 28, 31, 33 and 45. In a modified box plot, where does the upper whisker end?
Answer: A
Why: The IQR is 8, so the upper fence is 20 plus 12, which is 32. The values 33 and 45 lie beyond it and are drawn as separate points. The largest value not beyond the fence is 31, and the whisker ends there.
In context: the upper whisker reaches 31 minutes, and two unusually long commutes, 33 and 45 minutes, appear as separate points. The 33 is only just beyond the fence, a reminder that the label means look closer, not throw away.
If you chose the fence, compare the trap in Part 4: the fence is a test line, and Try It 2.25's whisker stopped at 240 with the fence at 252.5 for the same reason.
Explain it to yourself
The idea at the centre of this lesson is the one most people get backwards at first. Explaining it in your own words is the best test of whether it has stuck.
Discussion prompt
In the heights box plot, the second quarter spans 1.5 inches and the fourth spans 7 inches. A classmate says the fourth quarter must contain more students. In two or three sentences, explain why that is wrong and what the widths do tell you.
A strong answer says where the equal counts come from: the cuts are made by counting places. A very strong one adds a number, ten students in 1.5 inches is about 6.7 per inch, against about 1.4 per inch in the fourth quarter.
If your explanation used the word more, check what it was more of. More room and more students are different claims, and a box plot only ever shows the first.
The same reasoning works for every part of every box plot, including the whiskers, which is why it is worth putting into words once, carefully, now.
Exit ticket
A commuter records how many minutes late her bus is on eleven mornings: 0, 1, 1, 2, 3, 3, 4, 5, 6, 8 and 25. Answer from the data alone, then say what the plot would show.
Discussion prompt
Find the five-number summary and the fences, decide whether any value is a potential outlier, say where the whiskers of a modified box plot end, and describe the delays in one sentence.
With eleven values the median is the 6th value and each half holds five, so Q1 is the 3rd value and Q3 the 9th. If you got Q1 equal to 1.5, you kept the median in the lower half.
If you ended the whisker at 13.5, go back to the fence trap: the whisker stops at 8, a morning that actually happened.
And if you dropped the 25 from the data before finding the quartiles, put it back. A modified plot keeps the outlier in every calculation; it only changes how the value is drawn.
Recap
A box plot draws the five-number summary on a scaled number line. Because the quartiles cut the data by counting, each of its four parts holds about a quarter of the values, so the widths report concentration: narrow means crowded and wide means spread out, never more or fewer.
| idea | how to find or draw it | read it as |
|---|---|---|
| five-number summary | min, Q1, median, Q3, max | four parts of about a quarter each |
| median | place (n + 1)/2 in the sorted list | half the data on each side |
| Q1 and Q3 | medians of the lower and upper halves | a quarter below Q1, a quarter above Q3 |
| IQR | Q3 − Q1, the box width | spread of the middle half; resistant |
| range | max − min, the whole plot | spread of the extremes; not resistant |
| fences | Q1 − 1.5 IQR and Q3 + 1.5 IQR | beyond them: a potential outlier |
| modified plot | whiskers to the extreme values inside the fences | outliers drawn as separate points |
| parallel plots | several groups on one shared axis | compare medians, boxes, whiskers |
The habits that matter: count places, never measure the scale, when finding a quartile; draw the axis first; put the median where the data put it; end a whisker at a real value; and compare groups only on one number line, in sentences that name the feature and the units.
A box plot cannot show the sample size, clusters or gaps, so pair it with a histogram or a dot plot when those matter. Section 2.5 turns to the centre of the data, where the median you have used here meets the mean.
OpenStax Introductory Statistics 2e, §2.4 Box Plots §2.4, pp. 94-98 — the whole section
OpenStax Introductory Business Statistics 2e, §2.1 Display Data §2.1 Display Data — box plots in business contexts
Want this taught 1-on-1? Alexander tutors Statistics — $55/session, free consultation.