The first displays of the descriptive chapter. The stem-and-leaf plot, which splits each observation into a stem and a final significant digit and so keeps every original value while making the shape visible; reading a stemplot for concentration, gaps and outliers; the side-by-side stemplot that compares two data sets against one shared spine of stems; the line graph, whose joined points assert that the horizontal axis is ordered; and the bar graph for qualitative data, whose separated bars assert that it is not.
Subject: Statistics · 69 slides · symbolic lesson
Open the interactive version of this deck
Title
Statistics · Section 2.1
Three ways to turn a list of numbers into a picture you can read
Objectives
Chapter 1 was about collecting data well. Chapter 2 is about what you do once the numbers are in front of you, and the first answer is to draw them. A list of thirty scores tells you almost nothing at a glance; the right picture tells you the shape in a second.
This section builds three of those pictures by hand, and for each one you will learn not just the recipe but what the picture claims about the data. Every graph makes claims, and a graph drawn with the wrong recipe makes false ones.
The arithmetic all lesson is counting and dividing. The skill is judgement: choosing the picture, building it honestly, and then saying in a sentence what it shows about the people or things that were measured.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 65-66 — the chapter introduction
Warm-up
Here are the thirty-one first-exam scores from Professor Dean's pre-calculus class, already sorted from smallest to largest. They are the data for the first example of the section.
\[ 33, 42, 49, 49, 53, 55, 55, 61, 63, 67, 68, 68, 69, 69, 72, 73 \]
\[ 74, 78, 80, 83, 88, 88, 88, 90, 92, 94, 94, 94, 94, 96, 100 \]
Discussion prompt
Without drawing anything, answer three questions from the list. How many students scored in the 60s? How many scored 90 or more? In which ten points, such as the 50s or the 80s, do the most scores fall?
Notice how the effort felt. Every question meant a fresh pass through 31 numbers, and the list never got easier to read. By the end of Part 1 you will have a picture of this list that answers all three questions at a glance and still keeps every single score.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — Example 2.1 data
Section
Part 1
Concept
The book opens the chapter with a reason. A statistical graph is a tool for learning about the shape, or distribution, of a sample or a population, and a graph can be more effective than a mass of numbers because you can see where the data cluster and where there are only a few values.
Statisticians often graph data first, before computing anything, to get a picture of what they are working with. More formal tools come afterwards. The order matters: a mean or a percentage computed without looking can hide a typing error, a gap, or a value that does not belong.
The chapter names a whole family of graphs: dot plots, bar graphs, histograms, stem-and-leaf plots, frequency polygons, pie charts and box plots. This section takes three of them.
| graph | built for | what it keeps |
|---|---|---|
| stem-and-leaf plot | a small set of numbers | every original value, plus the shape |
| line graph | frequencies of a few ordered values | how often each value occurs, as a trend |
| bar graph | categories | the size of each category, side by side |
Keep the last column in mind. Each display throws something away in exchange for clarity, and choosing a display means choosing what you can afford to lose.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 65-66 — the introduction to Descriptive Statistics
Concept
The stem-and-leaf graph, or stemplot, comes from exploratory data analysis, and the book calls it a good choice when the data set is small. The whole idea is one cut made in every number.
stem and leaf — To build a stemplot, divide each observation into a stem and a leaf. The leaf consists of a final significant digit; the stem is everything in front of it.
Figure (svg): Four numbers written as rows of digit tiles. In 23 the tile 2 is blue and 3 is orange; in 432 the tiles 4 and 3 are blue and 2 orange; in 5,432 the tiles 5, 4, 3 are blue and 2 orange; in 9.3 the 9 is blue and the 3 after the decimal point is orange. Beside each, the stem and the leaf are written out.
The book's four examples show the rule from every side. 23 has stem 2 and leaf 3. 432 has stem 43 and leaf 2, and 5,432 has stem 543 and leaf 2, so the stem is not limited to one digit. The decimal 9.3 has stem 9 and leaf 3: the leaf can sit after the decimal point.
\[ 23 = 2 \times 10 + 3, \qquad 432 = 43 \times 10 + 2 \]
That line is why the cut works. Writing the leaf as the last digit is the same as writing the number as so many tens plus a remainder, so stem and leaf together lose nothing, and you can always rebuild the original value from them.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — the stem-and-leaf graph
Matching
Practise the cut before building a whole plot. For each value, the leaf is the single final digit and the stem is whatever is left in front of it. Watch the decimals and the three-digit values in particular.
Match the pairs
Why: In every case the leaf is one digit and the stem is all the rest. The score of 100 has stem 10, not 1, because cutting off the final zero leaves ten. The value 1,204 has stem 120 and leaf 4, not stem 12 and leaf 04, since a leaf is never two digits. And 0.7 has stem 0, which is a real stem that must be written.
The two slips that catch people are both about the leaf. Writing 1,204 as stem 12 with leaf 04 puts two digits in one leaf, which breaks the rule that one leaf stands for one value. Writing 100 under stem 1 confuses it with a score of 10.
A stem of zero, as for 0.7, is easy to forget because it feels like nothing. It is a row like any other, and the commute data later in this lesson needs it.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — the examples 23, 432, 5,432 and 9.3
Notation
A finished stemplot is a compact notation, and like any notation it has to be read correctly. Take one row from the exam plot apart.
Annotate
On: \( 6 \mid 1\;\;3\;\;7\;\;8\;\;8\;\;9\;\;9 \)
Two misreadings are common. One reads the row as the number 1,378,899, as if the leaves were a numeral; they are a list. The other reads the leaves as separate stems; they are not, because there is only one stem per row.
\[ 6 \mid 8 \;\Rightarrow\; 6 \times 10 + 8 = 68 \]
The key, written under every stemplot, settles what a stem and leaf mean in the units of the problem. Without a key, the row could equally be 6.1 kilometres or 610 dollars.
Concept
The book gives the recipe in three sentences: write the stems in a vertical line from smallest to largest, draw a vertical line to the right of the stems, then write the leaves in increasing order next to their stem. Each rule protects something.
\[ \text{key: } 4 \mid 2 = 42 \text{ points} \quad \text{or} \quad 4 \mid 2 = 4.2 \text{ km} \]
The first rule is the one people break, because an empty row looks like a mistake. It is the opposite. The rows are evenly spaced amounts, and leaving one out would pull the values on either side closer together than they really are.
The book's own exam data arrive already sorted, which makes ordering the leaves free. With unsorted data, a quick way is to file each value in its row as you meet it, then rewrite each row in order.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — constructing a stemplot
Concept
The rule says the leaf is the final significant digit, and for whole-number scores or one-decimal distances that is simply the last digit you see. Real data are sometimes recorded more finely than a stemplot can use, and then you decide what the leaf is before you start.
Suppose times were recorded to the hundredth of a second: 12.47, 12.83, 13.05 and so on. Taking the last digit as the leaf would make the stems 124, 128, 130, hundreds of rows for a spread of a couple of seconds, almost all of them empty. The plot would show no shape at all.
\[ 12.47 \to 12.5 \to 12 \mid 5 \]
The usual fix is to round each value to the precision you want to display, here to the tenth, and then cut. The stem becomes whole seconds and the leaf the tenths digit. Some books truncate instead of rounding, simply dropping the extra digit; either is acceptable if you say which in the key.
\[ \text{key: } 12 \mid 5 = 12.5 \text{ s (rounded to the nearest tenth)} \]
Rounding costs you the exact values, which is the stemplot's main advantage, so do it only when the plot is otherwise unreadable. The book's own data never need it: every example is already recorded at a precision that gives a sensible number of stems.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — the leaf consists of a final significant digit
Worked example
The book's first example takes the thirty-one exam scores from the warm-up and builds the stemplot. Work it one row at a time, keeping a running count so that nothing is lost.
Find the smallest and the largest score
Why: they decide where the column of stems starts and stops.
\[ 33 \to \text{stem } 3, \qquad 100 \to \text{stem } 10 \]
Write every stem from 3 to 10
Why: eight rows, one for each ten, in order, before a single leaf is written.
\[ 3, \; 4, \; 5, \; 6, \; 7, \; 8, \; 9, \; 10 \]
File the scores below 60
Why: each score's last digit goes in the row of its tens.
\[ 3 \mid 3 \qquad 4 \mid 2\;9\;9 \qquad 5 \mid 3\;5\;5 \]
Because the list is already sorted, each row's leaves arrive in increasing order on their own. With unsorted data you would file each score as you met it and tidy each row afterwards.
File the 60s and 70s
Why: the repeated 68 and 69 each keep their own leaf.
\[ 6 \mid 1\;3\;7\;8\;8\;9\;9 \qquad 7 \mid 2\;3\;4\;8 \]
File the 80s, the 90s and the 100
Why: cutting off the final zero of 100 leaves the stem 10.
\[ 8 \mid 0\;3\;8\;8\;8 \qquad 9 \mid 0\;2\;4\;4\;4\;4\;6 \qquad 10 \mid 0 \]
Figure (svg): A stemplot of the 31 exam scores with stems 3 to 10. The rows read: 3, leaf 3; 4, leaves 2 9 9; 5, leaves 3 5 5; 6, leaves 1 3 7 8 8 9 9; 7, leaves 2 3 4 8; 8, leaves 0 3 8 8 8; 9, leaves 0 2 4 4 4 4 6; 10, leaf 0. The rows for 9 and 10 are highlighted, with a note that 8 of the 31 scores are 90 or more.
Look at what the picture gives that the list did not. The two long rows, the 60s and the 90s, jump out. The single 33 sits alone at the bottom. And every score can still be read back exactly.
Check that no score was lost
Why: add the row lengths; they must come to the 31 students in the class.
\[ 1 + 3 + 3 + 7 + 4 + 5 + 7 + 1 = 31 \]
All thirty-one leaves are present, one per student. This count is worth doing every time you build a stemplot; a leaf dropped while copying is the most common error, and it is invisible unless you count.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — Example 2.1 and Table 2.1
Worked example
The book draws one conclusion from the plot: that most scores fell in the 60s, 70s, 80s and 90s, and that eight of the 31, approximately 26 percent, were in the 90s or 100, a fairly high number of As. Check both claims from the rows.
Count the leaves on stem 9
Why: each leaf is one student who scored in the 90s.
\[ 9 \mid 0\;2\;4\;4\;4\;4\;6 \;\Rightarrow\; 7 \text{ students} \]
Add the single leaf on stem 10
Why: the 100 is an A too.
\[ 7 + 1 = 8 \text{ students} \]
Divide by the whole class
Why: a share needs the total as its denominator, not the size of one row.
\[ \dfrac{8}{31} = 0.2581 \]
Turn the proportion into a percent
Why: multiply by one hundred.
\[ 0.2581 \times 100 = 25.8\% \]
The book rounds this to about 26 percent, and 25.8 does round to 26. So a quarter of the class, slightly more, earned an A on the first exam.
Count the 60s to 90s
Why: four rows, read off as their lengths.
\[ 7 + 4 + 5 + 7 = 23 \text{ of } 31 \]
Check the share by undoing the division
Why: multiplying the proportion back by the class size must return the count of eight.
\[ 0.2581 \times 31 = 8.0 \]
The second claim checks too: 23 of 31 is 74 percent of the class between 60 and 99, so most scores really did fall in those four tens. In context, Professor Dean's class mostly passed, and about one student in four earned an A.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — Example 2.1
Picture it
Here is the reason a stemplot shows shape and a list does not. Turn the plot a quarter turn, so the stems run along the bottom and the leaves pile upward.
Figure (svg): The exam stemplot rotated so each stem is a column along the bottom, labelled 30s through 100. The leaves are stacked upward in each column, and a dashed orange outline around each column is labelled with its count: 1, 3, 3, 7, 4, 5, 7 and 1.
Each pile of leaves is now a column, and the height of each column is the number of scores in that ten. Draw an outline around each pile and you have a bar chart of the tens, which Section 2.2 will call a histogram.
The leaves inside the bars are the bonus. An ordinary bar chart of the tens would tell you seven scores fell in the 90s; the stemplot also tells you they were 90, 92, 94, 94, 94, 94 and 96.
That is the trade the book means when it says a stemplot gives an exact picture of the data. It keeps every value and shows the shape at once, which is only possible because the data set is small. With three hundred scores the piles would be too tall to write, and a histogram would take over.
\[ \text{row length} = \text{count}, \quad \text{because every leaf is one digit wide} \]
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 67 — the stemplot gives an exact picture of the data
Ranking
A stemplot built in the wrong order still usually comes out right, but the order below is the one that prevents the two classic errors: a missing stem and a lost leaf. Arrange the steps.
Put in order
Order the steps for building a stemplot from a raw list of numbers.
Why: You must know the range before you can list the stems, and the stems must all be written before any leaves, or an empty ten is easy to skip. Leaves are filed as you meet them and then put in order within each row. Counting the leaves at the end catches any value dropped along the way, and the key tells a reader what the digits mean.
Writing every stem before any leaf is the step that does the most work. If you write stems only as you need them, a ten with no data never gets a row, and the gap it should show disappears from the picture.
The final count costs ten seconds. It is the only step that catches a value you skipped, and a skipped value changes every share you compute afterwards.
The key comes last only because it is the last thing written, not because it matters least. A stemplot handed to someone without its key is a page of digits they cannot turn back into data.
Worked example
The book's first Try It gives the scores of the Park City team's last thirty games, sorted, and asks for the stemplot. The book prints no worked answer, so build it here and check it by counting.
\[ 32, 32, 33, 34, 38, 40, 42, 42, 43, 44, 46, 47, 47, 48, 48 \]
\[ 48, 49, 50, 50, 51, 52, 52, 52, 53, 54, 56, 57, 57, 60, 61 \]
Read the range
Why: the smallest score is 32 and the largest 61, so the stems run from 3 to 6.
\[ 32 \to 3, \qquad 61 \to 6 \]
File the 30s
Why: five games.
\[ 3 \mid 2\;2\;3\;4\;8 \]
File the 40s
Why: twelve games, with three 48s each written out.
\[ 4 \mid 0\;2\;2\;3\;4\;6\;7\;7\;8\;8\;8\;9 \]
Count as you go. The 40s row is the longest so far, twelve leaves, and the next row will turn out to be almost as long, which is the first hint of where this team's scores centre.
File the 50s
Why: eleven games.
\[ 5 \mid 0\;0\;1\;2\;2\;2\;3\;4\;6\;7\;7 \]
File the 60s
Why: two games.
\[ 6 \mid 0\;1 \]
Figure (svg): A stemplot of 30 basketball scores with stems 3 to 6: 3, leaves 2 2 3 4 8; 4, leaves 0 2 2 3 4 6 7 7 8 8 8 9; 5, leaves 0 0 1 2 2 2 3 4 6 7 7; 6, leaves 0 1. The 4 and 5 rows are highlighted as nearly equal and much the longest.
The two middle rows carry almost everything, and they are almost the same length, so the team's scoring is centred near 50 points with few games far from it.
Check the total against the thirty games
Why: add the four row lengths.
\[ 5 + 12 + 11 + 2 = 30 \]
All thirty games are present. In context: in 23 of its last 30 games the Park City team scored between 40 and 59 points, and it never scored below 32 or above 61.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 67 — Try It 2.1
Estimation
A stemplot lets you estimate before you count, because row lengths are visible. Look at the Park City plot once more: the 50s row has eleven leaves and the 60s row has two.
Predict first
Roughly what share of the thirty games did the team score 50 points or more?
Correct: about two fifths
Why: Eleven games in the 50s plus two in the 60s is thirteen of thirty. Thirteen is a little under half of thirty, and two fifths of thirty is twelve, so two fifths is the closest. Exactly, thirteen divided by thirty is 0.433, or about 43 percent.
\[ \dfrac{11 + 2}{30} = \dfrac{13}{30} = 0.433 \]
The estimation habit is to compare row lengths with the total before reaching for a calculator. The 40s row alone is twelve, the 50s row and the 60s row together are thirteen, so the split above and below 50 is close to even, with slightly more games below.
If you picked three fifths, you probably added the 40s row. The 40s are below 50, even though the row sits next to the 50s on the page. Read the stems, not the position on the screen.
Fill the middle
One more reading from the Park City plot, done carefully. The busiest row is the 40s, with twelve leaves, and the question is what fraction of the team's games that row represents.
Fill in the blanks
\dfrac12}0.4 = ___
Why: The 40s row holds twelve leaves, one per game, and the team played thirty games, so the share is twelve over thirty, which is 0.4, or 40 percent.
Notice which number goes on the bottom. The denominator is all thirty games, not the twelve in the row and not the four rows of the plot. A share always says of what, and here it is of the team's last thirty games.
So two games in every five landed in the 40s. Together with the 50s row, those two rows hold 23 of 30 games, about 77 percent, which is the plot's main message in one number.
\[ \dfrac{12 + 11}{30} = \dfrac{23}{30} = 0.767 \]
Section
Part 2
Concept
The rule never changed: the leaf is the final significant digit. With data recorded to one decimal place, the final digit is the tenths digit, so the leaf sits after the decimal point and the stem is the whole-number part.
\[ 4.5 \to 4 \mid 5, \qquad 12.3 \to 12 \mid 3, \qquad 0.7 \to 0 \mid 7 \]
The book adds a note to its distance example for exactly this reason: the leaves are to the right of the decimal. Nothing in the plot itself shows a decimal point, so the key carries all the meaning.
\[ \text{key: } 4 \mid 5 = 4.5 \text{ km} \]
Here is the danger. Without a key, the row 4 bar 5 could be 45 kilometres, and a reader would have no way to know. The same digits on the page describe two data sets a factor of ten apart.
The stems are now whole kilometres, so each row covers one kilometre, from 4.0 up to 4.9. Every stem from the smallest to the largest must still be written, and with decimals that can mean many empty rows, as the next example shows. Those empty rows are where the most interesting thing in the data turns out to be.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 67 — Example 2.2 note: the leaves are to the right of the decimal
Worked example
The data are the distances, in kilometres, from one home to twenty-one local supermarkets. The book asks for a stemplot and then whether the data have any concentration of values.
\[ 1.1, 1.5, 2.3, 2.5, 2.7, 3.2, 3.3, 3.3, 3.5, 3.8, 4.0 \]
\[ 4.2, 4.5, 4.5, 4.7, 4.8, 5.5, 5.6, 6.5, 6.7, 12.3 \]
Choose the cut
Why: the stem is the whole number of kilometres, the leaf the tenths digit.
\[ 3.8 \to 3 \mid 8 \]
Write every stem from 1 to 12
Why: the smallest value is 1.1 and the largest 12.3, so twelve rows.
\[ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 \]
File stems 1 to 3
Why: five shops so far, then five more.
\[ 1 \mid 1\;5 \qquad 2 \mid 3\;5\;7 \qquad 3 \mid 2\;3\;3\;5\;8 \]
Watch the value 4.0. Its leaf is 0, and it is easy to skip a zero leaf as if it were nothing. It is a shop four kilometres away and it must be written.
File stems 4 to 6
Why: 4.0 has leaf 0, which is a real leaf.
\[ 4 \mid 0\;2\;5\;5\;7\;8 \qquad 5 \mid 5\;6 \qquad 6 \mid 5\;7 \]
Leave stems 7 to 11 empty and file 12.3
Why: no shop is between 7 and 12 kilometres away, and the rows say so.
\[ 7 \mid \quad 8 \mid \quad 9 \mid \quad 10 \mid \quad 11 \mid \quad 12 \mid 3 \]
Figure (svg): A stemplot of 21 supermarket distances in kilometres, stems 1 to 12. Stems 1 to 6 carry leaves: 1 5; 3 5 7; 2 3 3 5 8; 0 2 5 5 7 8; 5 6; 5 7. Stems 7, 8, 9, 10 and 11 are written with no leaves, and stem 12 carries the single leaf 3, drawn in red. Stems 3 and 4 are highlighted as the concentration.
Check the count
Why: add the row lengths, empty rows contributing zero.
\[ 2 + 3 + 5 + 6 + 2 + 2 + 0 + 0 + 0 + 0 + 0 + 1 = 21 \]
All twenty-one shops are filed. The book's answer to its question follows straight from the picture: values concentrate at three and four kilometres, and the value 12.3 may be an outlier.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 67-68 — Example 2.2 and Table 2.2
Trap
A student builds the supermarket plot and, to save space, writes only the stems that have leaves.
\[ 5 \mid 5\;6 \]
\[ 6 \mid 5\;7 \]
\[ 12 \mid 3 \]
On the page, 12.3 now sits one row below 6.7, like any other neighbour. The plot looks tidy and smooth.
The stems are a number line on end. A missing row is a missing stretch of that line.
\[ 6 \mid 5\;7 \]
\[ 7 \mid \quad \dots \quad 11 \mid \]
\[ 12 \mid 3 \]
Five empty rows put 12.3 where it belongs: far from the rest. The empty rows are the evidence for the outlier, so they must stay.
Concept
The book says what to look for in any stemplot: an overall pattern and any outliers. The pattern is where the values crowd together and how they spread out; the outlier is the value that refuses to join the pattern.
outlier — An observation of data that does not fit the rest of the data. It is sometimes called an extreme value. When you graph an outlier, it will appear not to fit the pattern of the graph.
Figure (svg): A dot plot of the 21 distances on a kilometre axis from 0 to 13. Twenty dots crowd between 1.1 and 6.7, stacked where values repeat, and one red dot sits alone at 12.3. A yellow bracket spans the empty stretch from 6.7 to 12.3, labelled 5.6 km.
Put the supermarket distances on an ordinary number line and the definition becomes visible. Twenty shops sit together between 1.1 and 6.7 kilometres. One shop sits alone at 12.3, separated from the crowd by an empty stretch of 5.6 kilometres.
Notice too what the stemplot and the dot plot share. Both keep every distance on a true scale with nothing skipped, so a gap in the data becomes a gap on the page. Any display that squeezes out empty space would hide this shop in plain sight.
Notice what makes it an outlier. It is not simply the largest value, since every data set has a largest value. It is the largest value with a gap in front of it that is far bigger than the ordinary spacing between neighbours. The next slide measures that spacing.
The book says outliers will be covered in more detail later, because explaining one takes background knowledge. Chapter 2.3 gives a numerical rule. For now the test is visual and honest: does the value join the pattern or not?
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 67 — overall pattern and outliers
Worked example
The book's solution says only that 12.3 may be an outlier. Make that judgement with numbers rather than with an impression, by measuring how far apart neighbouring values are.
List the gaps between neighbouring values
Why: subtract each sorted distance from the next; nineteen of the twenty gaps are small.
\[ 0.4, 0.8, 0.2, 0.2, 0.5, 0.1, 0, 0.2, 0.3, 0.2, \dots \]
Find the largest ordinary gap
Why: among the first nineteen, the biggest step is from 5.6 to 6.5.
\[ 6.5 - 5.6 = 0.9 \text{ km} \]
Measure the gap in front of 12.3
Why: from the next largest value, 6.7.
\[ 12.3 - 6.7 = 5.6 \text{ km} \]
Compare the two gaps
Why: divide to see how many ordinary gaps fit in the big one.
\[ \dfrac{5.6}{0.9} = 6.2 \]
Figure (svg): A lollipop chart of the 20 gaps between neighbouring sorted distances, numbered 1 to 20 along the bottom. Nineteen gaps are at most 0.9 km tall; the twentieth, the gap from 6.7 to 12.3, rises to 5.6 km and is drawn in red.
So the jump to 12.3 is more than six times as big as any other step in the data. That is what not fitting the pattern means here.
Check against the spread of the rest
Why: the other twenty values run from 1.1 to 6.7; their whole spread equals the single gap in front of 12.3.
\[ 6.7 - 1.1 = 5.6 = 12.3 - 6.7 \]
One empty stretch as wide as the entire rest of the data set. In context: twenty of the shops are within about seven kilometres of this home, and one is twice that far away, worth checking before it is trusted or used.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 67-68 — Example 2.2 solution
Sorting
Every data set has a largest value, but only some have an outlier. For each situation, decide whether the value named stands apart from the pattern or simply sits at its end. Ask about the gap, not the size.
Sort into buckets
Sort each named value: an outlier, or just the end of the pattern?
The exam score of 33 is the honest borderline case. A gap of nine is the biggest in that data, yet the scores have other gaps of seven and six, so 33 does not stand apart the way 12.3 does. Reasonable people could flag it; the important thing is that you can say what you compared.
The height case is the book's other warning in disguise. Some outliers are mistakes, like writing 50 instead of 500, and the first thing to do with a startling value is to check how it was recorded.
Concept
Finding an outlier is one part of reading a stemplot. The other part is describing the overall pattern, and a good description answers three questions about the rows that do carry the data.
Apply them to the supermarket distances without the 12.3. The crowd is at three and four kilometres, eleven of the twenty-one shops. The spread runs from 1.1 to 6.7 kilometres, six stems. And the pile is fairly even, with two or three shops on each side of the crowd, thinning at a similar rate.
\[ \dfrac{11}{21} = 0.524 \text{ of the shops at 3.0 to 4.9 km} \]
Compare the exam scores. They crowd in the 60s and 90s, spread over eight stems from 33 to 100, and trail off further at the low end, where a few rows of one or three leaves stretch down to 33. A pile that trails off on one side like that is called skewed, a word Section 2.6 makes precise.
Always finish the description in the language of the problem: shops and kilometres, students and points. A description that never mentions what was measured has not described the data.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 67 — look for an overall pattern and any outliers
Worked example
Thirty off-campus statistics students report how far, in miles, they live from the college. The book asks for a stemplot and any outliers, and prints no worked answer.
\[ 0.5, 0.7, 1.1, 1.2, 1.2, 1.3, 1.3, 1.5, 1.5, 1.7, 1.7, 1.8, 1.9, 2.0, 2.2 \]
\[ 2.5, 2.6, 2.8, 2.8, 2.8, 3.5, 3.8, 4.4, 4.8, 4.9, 5.2, 5.5, 5.7, 5.8, 8.0 \]
Choose the cut and the stems
Why: whole miles for stems, tenths for leaves; the values run from 0.5 to 8.0, so stems 0 to 8.
\[ 0, 1, 2, 3, 4, 5, 6, 7, 8 \]
File stems 0 to 2
Why: the stem 0 holds the two students under a mile.
\[ 0 \mid 5\;7 \qquad 1 \mid 1\;2\;2\;3\;3\;5\;5\;7\;7\;8\;9 \qquad 2 \mid 0\;2\;5\;6\;8\;8\;8 \]
Already the shape is showing. Eleven students sit on stem 1, between one and two miles, which is more than a third of the class on a single row.
File stems 3 to 5
Why: the rows thin out.
\[ 3 \mid 5\;8 \qquad 4 \mid 4\;8\;9 \qquad 5 \mid 2\;5\;7\;8 \]
Leave 6 and 7 empty and file 8.0
Why: the leaf of 8.0 is 0.
\[ 6 \mid \qquad 7 \mid \qquad 8 \mid 0 \]
Figure (svg): A stemplot of 30 commute distances in miles with stems 0 to 8. Rows: 0, leaves 5 7; 1, leaves 1 2 2 3 3 5 5 7 7 8 9; 2, leaves 0 2 5 6 8 8 8; 3, leaves 5 8; 4, leaves 4 8 9; 5, leaves 2 5 7 8; 6 and 7 empty; 8, leaf 0 in red.
Measure the gap before 8.0
Why: and compare it with the largest other gap, from 2.8 to 3.5.
\[ 8.0 - 5.8 = 2.2, \qquad 3.5 - 2.8 = 0.7 \]
Check the count and the verdict
Why: the rows must hold thirty students, and the gap ratio says how unusual 8.0 is.
\[ 2 + 11 + 7 + 2 + 3 + 4 + 0 + 0 + 1 = 30, \qquad \dfrac{2.2}{0.7} = 3.1 \]
All thirty students are filed. The 8.0-mile student sits beyond a gap more than three times any other, so 8.0 is a reasonable candidate for an outlier, though a far milder one than the 12.3-kilometre supermarket. In context, most students live within about six miles, crowded between one and three.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 68 — Try It 2.2
Socratic
Finding an outlier is the start of a question, not the end of one. The book gives two possible origins, and the right response depends entirely on which one you are facing.
Discussion prompt
Give two different explanations for the supermarket at 12.3 kilometres, one of each kind the book describes. For each, say what you would do with the value.
The response that is never right is to delete the value because it spoils the picture. An outlier that is genuine may be the most important observation in the set, and deleting it hides exactly what someone needed to know.
So the stemplot's job is to make the value visible. Your job is to go and find out why it is there.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 67 — some outliers are due to mistakes, others indicate something unusual
Trap
A student looks at the exam stemplot and writes:
\[ 10 \mid 0 \;\Rightarrow\; \text{outlier} \]
The reasoning is that 100 is the most extreme score, it sits alone on its stem, and outliers are extreme values.
A lone leaf on the top row feels like a value set apart.
Measure the gap instead of reading the row.
\[ 100 - 96 = 4 \]
\[ \text{other steps: up to } 9 \]
A step of four is ordinary in these scores. The 100 is the end of the pattern, not a break from it.
A lone leaf on a stem says only that the value is in its own ten. Only empty stems show a gap.
Section
Part 3
Notation
A side-by-side stem-and-leaf plot compares two data sets in two columns. Two sets of leaves share the same stem: one set to the left of the stems and one to the right. Take one row of the book's president plot apart.
Annotate
On: \( 9\;9\;8\;7\;7\;7\;6\;3\;2 \;\mid\; 4 \;\mid\; 6\;9 \)
The left side's ordering is the convention people get backwards. On both sides, the smallest leaf is nearest the stem and the leaves grow outward, so the left leaves look reversed when you read along the line.
\[ \text{left: } 42, 43, 46, \dots \quad \text{written as } 9\;9\;8\;\dots\;3\;2 \]
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 68 — the side-by-side stem-and-leaf plot
Concept
You could draw two separate stemplots and set them side by side on the page. The side-by-side plot does something better: it forces the two data sets onto the same stems, so each row compares like with like.
Reading across one row answers a direct question. In the forties row, nine inaugurations and two deaths; in the sixties row, ten inaugurations and twelve deaths. Each row is a fair comparison because both sides cover exactly the same ten years.
\[ \text{stem } 6: \quad 10 \text{ inaugurations}, \quad 12 \text{ deaths} \]
Reading down each side answers the other question: where does each data set sit, and how spread out is it? A side that fills three rows is tightly bunched; a side that fills six rows is spread out.
Two rules follow from sharing one spine. The stems must cover the range of both data sets together, so one side may have empty rows the other side needs. And the key applies to both sides at once, so the two sets must be in the same units. You cannot put ages in years beside heights in centimetres on one spine.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 68 — two sets of leaves share the same stem
Worked example
Table 2.3 lists the ages of forty-four presidential inaugurations, from Washington at 57 to Obama at 47. Cleveland appears twice, at 47 and 55, because he served two separate terms and was inaugurated twice.
Find the range of both data sets together
Why: inauguration ages run from 42 to 69 and death ages from 46 to 93, so the shared stems run from 4 to 9.
\[ \min = 42, \quad \max = 93 \;\Rightarrow\; \text{stems } 4 \text{ to } 9 \]
File the inaugurations in the forties
Why: nine ages; the leaves increase outward, so they are written right to left.
\[ 42, 43, 46, 47, 47, 47, 48, 49, 49 \to 9\;9\;8\;7\;7\;7\;6\;3\;2 \mid 4 \]
Count the fifties
Why: twenty-five presidents took office in their fifties, the busiest row in the whole plot.
\[ 5 \mid 0\;1\;1\;1\;1\;1\;2\;2\;4\;4\;4\;4\;4\;5\;5\;5\;5\;6\;6\;6\;7\;7\;7\;7\;8 \]
On the plot these twenty-five leaves are written in the reverse order, leftward from the spine. The display above lists them smallest first so you can check them against the table.
Twenty-five leaves on one row is a lot to write, and it is where copying errors happen. Counting the fifties in the table first, then checking the row has that many leaves, is the safest way through.
File the sixties
Why: ten ages, from Truman at 60 to Reagan at 69.
\[ 60, 61, 61, 61, 62, 64, 64, 65, 68, 69 \to 9\;8\;5\;4\;4\;2\;1\;1\;1\;0 \mid 6 \]
Note the empty rows on this side
Why: no president was inaugurated at 70 or older in these data, so stems 7, 8 and 9 carry no left leaves.
\[ 7 \mid, \quad 8 \mid, \quad 9 \mid \text{ (left side empty)} \]
Check the left side's total
Why: the three rows must hold all forty-four inaugurations.
\[ 9 + 25 + 10 = 44 \]
The left side matches the book's Table 2.5 leaf for leaf. In context, every inauguration in the table happened between ages 42 and 69, and more than half of them in a single decade, the fifties.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 68-70 — Example 2.3, Tables 2.3 and 2.5
Worked example
Table 2.4 lists thirty-nine ages at death, for the presidents who had died when the table was compiled. Cleveland's 71 appears twice again, once for each term, so the table counts one death twice; the book keeps both, and so does this plot, but it is worth knowing.
File the forties and fifties
Why: Kennedy at 46, Garfield at 49, then five deaths in the fifties.
\[ 4 \mid 6\;9 \qquad 5 \mid 3\;6\;6\;7\;8 \]
File the sixties
Why: twelve deaths, the longest row on this side.
\[ 6 \mid 0\;0\;3\;3\;4\;4\;5\;6\;7\;7\;7\;8 \]
The sixties row on this side is almost as long as the fifties row was on the other. Already you can see the whole death distribution sitting one or two stems further down the spine.
File the seventies
Why: eleven, including the doubled 71.
\[ 7 \mid 0\;1\;1\;1\;2\;3\;4\;7\;8\;8\;9 \]
File the eighties and nineties
Why: five and four deaths.
\[ 8 \mid 0\;1\;3\;5\;8 \qquad 9 \mid 0\;0\;3\;3 \]
Figure (svg): A side-by-side stemplot with stems 4 to 9 down the middle. Ages at inauguration on the left: stem 4, nine leaves; stem 5, twenty-five leaves; stem 6, ten leaves; stems 7 to 9 empty. Ages at death on the right: stem 4, leaves 6 9; stem 5, leaves 3 6 6 7 8; stem 6, twelve leaves; stem 7, eleven leaves; stem 8, leaves 0 1 3 5 8; stem 9, leaves 0 0 3 3.
Compare the stem 5 row with the book's Table 2.5. The book prints the leaves 3, 6, 6, 7, 7, 8, which would mean two presidents died at 57. Table 2.4 has only one: Harding, at 57. The correct row has five leaves, as drawn here.
Check the right side's total
Why: the six rows must hold the thirty-nine listed deaths, and with the extra 7 they would not.
\[ 2 + 5 + 12 + 11 + 5 + 4 = 39 \]
Thirty-nine leaves for thirty-nine table entries, so the corrected row is right and the printed plot has forty leaves, one too many. In context: the listed presidents died at ages from 46 to 93, most often in their sixties and seventies.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 69-70 — Example 2.3, Tables 2.4 and 2.5
Worked example
The book builds the plot and stops. A plot that is never read is half a worked example, so read this one, using only counts you can take from the rows.
Find each side's busiest row
Why: the longest row on each side shows where that data set concentrates.
\[ \text{inauguration: } 5 \mid (25) \qquad \text{death: } 6 \mid (12) \]
Find the middle value of the inauguration ages
Why: with 44 values the middle falls between the 22nd and 23rd; nine leaves come before stem 5, so these are its 13th and 14th leaves.
\[ \dfrac{54 + 55}{2} = 54.5 \]
Find the middle value of the death ages
Why: with 39 values it is the 20th; 2 + 5 + 12 = 19 come before stem 7, so it is that row's first leaf.
\[ 20\text{th value} = 70 \]
Figure (svg): Two dot plots on one axis of age in years, 40 to 95. Upper, in orange, the 44 ages at inauguration from 42 to 69, stacked up to five high, with the middle value 54.5 marked. Lower, in blue, the 39 ages at death from 46 to 93, with the middle value 70 marked.
The dot plots draw the same comparison on one shared ruler. The death ages sit to the right of the inauguration ages, as they must, and they are spread across a much wider stretch.
Compare the spreads
Why: count the rows each side occupies and its range.
\[ \text{inauguration: } 69 - 42 = 27, \qquad \text{death: } 93 - 46 = 47 \]
Check the middle values by counting from the other end
Why: for the death ages, the rows above stem 7 hold 5 + 4 = 9, and the stem 7 row holds 11, so 20 values are 70 or older and 19 are younger, which puts 70 exactly in the middle.
\[ 19 + 1 + 19 = 39 \]
Section 2.3 will call these middle values medians. In context: a typical president took office at about 54 and died at about 70, roughly fifteen years apart, and ages at death are spread far more widely than ages at inauguration.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 70 — Example 2.3 solution, Table 2.5
Error analysis
Four students start the side-by-side plot of the Atlanta Hawks' wins and losses, and each one writes a row that is wrong in a different way. Find each fault before you open the notes.
Annotate
On: \( 2\;5\;5\;6\;9 \mid 2 \mid 9\;8\;8\;6\;5 \quad 1 \mid 3 \quad 4\;6 \mid 4 \)
\[ \text{correct: } 9\;6\;5\;5 \mid 2 \mid 5\;6\;8\;8\;9 \]
The correct row is above. Both sides keep their smallest leaf beside the spine, which is what makes the two sides mirror each other on the page.
All four errors come from forgetting one sentence of the rules: one digit per leaf, smallest nearest the stem, both data sets on every stem. The last error is the most serious, since it silently changes values.
Step zero
Try It 2.3 gives the Atlanta Hawks' wins and losses for forty-two seasons and asks for a side-by-side stemplot. Before filing a single leaf, there are decisions to make and a check worth running on the data themselves.
Discussion prompt
Before you build the plot, what range of stems do you need, which data set goes on which side, and what simple check can you run on each season's pair of numbers?
Run that check on all forty-two seasons and every one adds to 82 except season 31, which has 19 losses and 31 wins, 50 games. That was a shortened lockout season, not a typing error, and it is a good example of checking data before trusting it.
A check like this costs a minute and can catch a transposed digit that no graph would reveal.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 70-71 — Try It 2.3, Table 2.6
Worked example
With the stems fixed at 1 to 6, file the forty-two loss totals on the left and the forty-two win totals on the right. The book prints no answer; the counts below were checked in node.
File stems 1 and 2
Why: one season each at the extremes, then the twenties.
\[ 9 \mid 1 \mid 3 \qquad 9\;6\;5\;5 \mid 2 \mid 5\;6\;8\;8\;9 \]
File stem 3
Why: fourteen losing totals and thirteen winning totals.
\[ 9\;9\;9\;6\;6\;6\;5\;4\;4\;2\;2\;2\;2\;0 \mid 3 \mid 0\;1\;1\;1\;1\;3\;4\;5\;5\;6\;6\;7\;8 \]
On the left side the leaves are written outward from the spine, so the row reads in decreasing order along the line. The thirty-nines sit right next to the stem's left edge and the thirties far out.
File stem 4
Why: thirteen losses and fourteen wins.
\[ 9\;8\;7\;7\;6\;6\;5\;4\;2\;1\;1\;0\;0 \mid 4 \mid 0\;1\;1\;2\;2\;3\;3\;3\;6\;6\;6\;7\;8\;8 \]
File stems 5 and 6
Why: the wins side has no season of 60 or more.
\[ 7\;6\;4\;4\;3\;2\;1\;1\;1 \mid 5 \mid 0\;0\;0\;0\;2\;3\;6\;7\;7 \qquad 9 \mid 6 \mid \]
Figure (svg): A side-by-side stemplot of the Atlanta Hawks' 42 seasons, stems 1 to 6. Losses on the left: stem 1, leaf 9; stem 2, leaves 5 5 6 9; stem 3, fourteen leaves; stem 4, thirteen leaves; stem 5, nine leaves; stem 6, leaf 9. Wins on the right: stem 1, leaf 3; stem 2, leaves 5 6 8 8 9; stem 3, thirteen leaves; stem 4, fourteen leaves; stem 5, leaves 0 0 0 0 2 3 6 7 7; stem 6 empty.
The two sides have almost the same shape, concentrated in the thirties and forties. The losses reach one row further, to the 69-loss season, and the wins reach one row lower, to the 13-win season, which is the same season seen from both sides.
Check both sides against forty-two seasons
Why: each side's rows must add to 42.
\[ \text{losses: } 1 + 4 + 14 + 13 + 9 + 1 = 42, \quad \text{wins: } 1 + 5 + 13 + 14 + 9 + 0 = 42 \]
Both sides are complete. In context, over these forty-two seasons the Hawks most often won and lost between 30 and 49 games, with 21 winning seasons, 19 losing seasons and 2 exactly even.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 70-71 — Try It 2.3
Picture it
The Hawks plot looks almost symmetric about its spine, and there is a reason that has nothing to do with basketball. In an 82-game season, every game is either won or lost.
\[ \text{wins} + \text{losses} = 82 \;\Rightarrow\; \text{losses} = 82 - \text{wins} \]
Figure (svg): Two rows of dots on a games axis from 10 to 70: the 42 win totals in blue on the upper row and the 42 loss totals in orange on the lower row, with a yellow dashed line at 41. Season 37 is joined by a line from 13 wins to 69 losses, and season 31, the 50-game season, is circled.
So each season's loss total is its win total reflected across 41, half of 82. A season with 13 wins is automatically a season with 69 losses; a season with 50 wins is automatically one with 32 losses. Draw the two data sets on one number line and every blue dot has an orange partner the same distance from 41 on the other side.
The one exception is the yellow pair, season 31, with 31 wins and 19 losses in a 50-game season. It is the only season whose pair is not mirrored around 41, because its total is not 82.
The lesson generalises. When two data sets are tied together by a fixed total, a side-by-side plot of them will always look like a reflection, and the comparison is telling you about one variable, not two.
Section
Part 4
Concept
Stemplots work when you want every individual value. Often the data take only a few values, repeated many times, and what you want is how often each one occurs. That is a frequency table, and a line graph draws it.
line graph — A graph in which the horizontal axis consists of data values and the vertical axis consists of frequency points. The frequency points are connected using line segments.
The book's example surveys forty parents on how many times per week a teenager must be reminded to do chores. Six possible answers, from 0 to 5 reminders, and the table records how many parents gave each one.
| times reminded per week | 0 | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|---|
| number of parents | 2 | 5 | 8 | 14 | 7 | 4 |
To draw it, you put the data values, 0 through 5, on the horizontal axis in order and evenly spaced. Above each value you plot a point at its frequency. Then you join neighbouring points with straight segments.
\[ (0, 2), \; (1, 5), \; (2, 8), \; (3, 14), \; (4, 7), \; (5, 4) \]
Before drawing anything, check the table adds up. Forty parents were surveyed, so the frequencies must total forty; if they do not, a row was miscopied.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 71 — line graphs, Table 2.7
Worked example
Draw the line graph for the forty parents' answers, then read it. The book shows the finished graph as Figure 2.2; here it is built and interpreted.
Check the frequencies add to the sample size
Why: every parent gave exactly one answer.
\[ 2 + 5 + 8 + 14 + 7 + 4 = 40 \]
Lay out the horizontal axis
Why: the data values in order, equally spaced, because one reminder more is the same step everywhere.
\[ 0, \; 1, \; 2, \; 3, \; 4, \; 5 \]
Plot each frequency above its value
Why: the highest frequency is 14, so a vertical axis to 16 holds everything.
\[ (3, 14) \text{ is the peak} \]
Label both axes with what they mean in this survey, times reminded per week and number of parents, not just x and y. A reader who meets the graph alone should know what was counted.
Join neighbouring points with segments
Why: the line lets the eye follow the rise and fall from one value to the next.
\[ (0,2) \to (1,5) \to (2,8) \to (3,14) \to (4,7) \to (5,4) \]
Figure (svg): A line graph with the number of times a teenager is reminded, 0 to 5, across the bottom and frequency, 0 to 16, up the side. Points at 2, 5, 8, 14, 7 and 4 are joined by straight segments, rising to a peak of 14 at three reminders and falling after it.
Turn the peak into a share
Why: fourteen of the forty parents gave the most common answer.
\[ \dfrac{14}{40} = 0.35 \]
Check the reading from both sides of the peak
Why: the parents below three reminders and above it, plus the fourteen at three, must again make forty.
\[ (2 + 5 + 8) + 14 + (7 + 4) = 15 + 14 + 11 = 40 \]
In context: the most common answer, from 35 percent of the parents, is three reminders a week, and slightly more parents report fewer reminders, 15, than more, 11. Only two parents said their teenager never needs reminding.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 71-72 — Example 2.4, Figure 2.2
Concept
A line graph can show counts or shares. Divide each frequency by the total and you have the relative frequency, the fraction of the sample that gave each answer.
\[ \text{relative frequency} = \dfrac{\text{frequency}}{\text{total}}, \qquad \dfrac{14}{40} = 0.35 \]
Figure (svg): Two line graphs of the reminder data side by side with identical shapes. The left panel's vertical axis is frequency, 0 to 16; the right panel's is relative frequency, 0 to 0.40, with the peak at 0.35.
Every point moves by the same factor, one fortieth, so the shape of the graph does not change at all. Only the numbers on the vertical axis do. The peak is still at three reminders, and every rise and fall is still in the same place.
Relative frequency earns its keep when you compare two surveys of different sizes. If another town asked 200 parents, its counts would be about five times as big and its line would tower over this one. Put both on a relative-frequency axis and they become directly comparable, each point a share of its own town.
\[ 0.05 + 0.125 + 0.2 + 0.35 + 0.175 + 0.1 = 1 \]
The relative frequencies always add to one, the whole sample, which is a quick check that the divisions were done right. Section 2.2 uses exactly this idea for histograms.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.2, p. 75 — relative frequency, previewed
Concept
The segments in a line graph are not decoration. Joining two points says that the values on the horizontal axis come in a meaningful order, so that moving from one point to the next is moving along a scale.
For the reminder data that claim is true. Zero, one, two, three, four and five reminders are ordered, each one more than the last, and they are equally spaced. So it makes sense to say the frequency rises from two reminders to three and falls from three to four. The line traces that rise and fall.
Be careful not to read more into a segment than that. The segment between two and three reminders does not mean some parents reported two and a half reminders; nobody could. It is a visual guide joining two real points, not a claim about values in between.
\[ \text{segment from } (2, 8) \text{ to } (3, 14): \quad \text{order, not in-between data} \]
When the horizontal axis has no natural order, the claim is false, and the line draws a trend that does not exist. The next slide shows how badly that can go.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 71 — the frequency points are connected using line segments
Prediction
Blood type is a category: O, A, B or AB. Suppose someone draws a line graph of the approximate share of Americans with each type, about 45, 40, 11 and 4 percent, in that order, and the line falls steadily. Now they redraw it with the types in a different order.
Predict first
If the types are listed AB, O, B, A instead, what happens to the line?
Correct: It changes shape completely, because the order of the categories was arbitrary
Why: The line follows whatever order the categories are written in. Listed AB, O, B, A, the points are 4, 45, 11 and 40, so the line shoots up, falls and rises again. Since nothing about blood types makes one order correct, neither the steady fall nor the zigzag means anything.
Figure (svg): Two line graphs of the same approximate blood-type percentages, O 45, A 40, B 11 and AB 4. On the left the types are ordered O, A, B, AB and the line falls steadily. On the right they are ordered AB, O, B, A and the line zigzags up, down and up again.
Look at the two panels together. Same four percentages, same axes, same scale, and yet one line says the shares fall steadily while the other says they swing up and down twice.
This is the whole case against a line graph for categories. A line's shape is supposed to be information. Here the shape is decided by whoever chose the order of the labels, so it is information about the person drawing the graph, not about blood.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 71-72 — line graphs for data values on the horizontal axis
Trap
A student draws favourite fruit for 30 classmates as a line graph.
\[ \text{apple } 12 \to \text{banana } 9 \to \text{cherry } 6 \]
The line falls steadily, and the write-up says preference declines across the fruits.
Declines from what to what? There is no scale here.
Ask first: is there a natural order on the horizontal axis?
\[ \text{fruit} \;\Rightarrow\; \text{no order} \]
\[ \text{no order} \;\Rightarrow\; \text{bar graph} \]
Separate bars show three sizes with no trend implied. Alphabetical order is a filing system, not a scale.
Worked example
Forty people were asked how many times in a year they had their car in the shop for repairs, and the book asks for a line graph. The answers run from 0 to 3, a counted and ordered variable, so a line is allowed.
| times in the shop | 0 | 1 | 2 | 3 |
|---|---|---|---|---|
| number of people | 7 | 10 | 14 | 9 |
Check the frequencies against the sample
Why: forty people, one answer each.
\[ 7 + 10 + 14 + 9 = 40 \]
Lay out the axis and plot the points
Why: four equally spaced values, each with its frequency above it.
\[ (0, 7), \; (1, 10), \; (2, 14), \; (3, 9) \]
Join neighbouring points
Why: three segments, rising to the peak at two visits and falling after it.
\[ (0,7) \to (1,10) \to (2,14) \to (3,9) \]
Figure (svg): A line graph of the number of times a car was in the shop in a year, 0 to 3, against frequency, 0 to 16. Points at 7, 10, 14 and 9 are joined, rising to a peak of 14 at two visits.
The line climbs for two steps and then drops, so the most common experience is in the middle of the range rather than at either end. Very few drivers avoided the shop entirely, and fewer than a quarter went three times.
Read the peak as a share
Why: fourteen of forty people.
\[ \dfrac{14}{40} = 0.35 \]
Check the share of people who needed any repair
Why: everyone except the seven with no visits, counted directly and by subtraction.
\[ 10 + 14 + 9 = 33 = 40 - 7, \qquad \dfrac{33}{40} = 0.825 \]
In context: the most common experience among these forty drivers was two shop visits in the year, reported by 35 percent, and 82.5 percent had their car in the shop at least once.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 72 — Try It 2.4, Table 2.8
Sorting
For each variable, ask the single question from this part: does the horizontal axis have a natural order with equal steps? If yes, a line graph of frequencies is allowed. If the values are categories with no order, use a bar graph.
Sort into buckets
Sort each variable by the graph that suits its frequencies.
The counted variables all begin with the words the number of, the same hint Chapter 1 gave for discrete data. A count has a built-in order and equal steps, which is exactly what a line graph needs.
Blood type deserves a second look because its labels are letters that seem to have an order, A before B. That order is the alphabet's, not the blood's; nothing about type A is less than type B, which is why the earlier figure could reshuffle them freely.
A bar graph is never wrong for counted data, just less informative about trend. A line graph is wrong for categories. When in doubt, the bars are the safe choice.
Section
Part 5
Concept
The book defines bar graphs by one visible feature: they consist of bars that are separated from each other. The bars can be rectangles, or rectangular boxes in three-dimensional plots, and they can be vertical or horizontal.
bar graph — A graph of bars separated from each other, one bar per category, whose lengths show the number or percentage in each category.
The separation is the message. Touching bars, as in a histogram, say the horizontal axis is a continuous number line, so one bar runs straight into the next. Gaps between bars say the categories are distinct, with nothing in between them.
Three consequences follow, and each is a rule worth keeping. The bars may be put in any order, since the categories have none, so choose the order that helps the reader, often largest first. Bars may run vertically or horizontally. And every bar must start at zero, because the length of a bar is what the eye compares.
\[ \text{bar length} \propto \text{value} \quad \text{only if the axis starts at } 0 \]
That last rule is the one most often broken in newspapers and advertising, and Part 5 ends by measuring what breaking it does.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 72 — bar graphs
Notation
Most of the book's bar graphs show proportions rather than counts, so every bar is a share of one named whole. Read the formula before using it on real tables.
Annotate
On: \( \text{percent} = \dfrac{\text{count in the category}}{\text{total of all categories}} \times 100 \)
The denominator is the place to slow down. A percent with the wrong total looks exactly like a correct one, and only the check that the shares add to 100 will catch it.
\[ \dfrac{1253540}{3215830} \times 100 = 38.98 \approx 39.0\% \]
Percentages let bars from groups of different sizes share one axis. A bar graph of counts would make any comparison between a big town and a small town a comparison of their sizes.
Worked example
Table 2.9 gives the percentage of U.S.-based TikTok users in five age groups, and the book asks for a bar graph. The book's graph has age groups on the horizontal axis and proportions on the vertical axis.
| age group | 10-19 | 20-29 | 30-39 | 40-49 | 50+ |
|---|---|---|---|---|---|
| percent of users | 32.5 | 29.5 | 16.4 | 13.9 | 7.1 |
Add the percentages before drawing
Why: shares of one whole should total 100.
\[ 32.5 + 29.5 + 16.4 + 13.9 + 7.1 = 99.4 \]
They total 99.4, not 100. The missing 0.6 percent is most likely rounding in the source or users outside these groups, and the book does not say which. A bar graph is still fine, because each bar stands on its own; it is a reason not to draw this as a pie.
Choose the axes
Why: age groups across, percent up, with the axis starting at 0 and running past the largest value.
\[ 0\% \text{ to } 35\% \]
The axis starts at zero, not at the smallest value, because the bars will be compared by length. Running it a little past 32.5 leaves room for the value labels above the tallest bar.
Draw one separated bar per group
Why: each bar's height is its percentage.
\[ 32.5, \; 29.5, \; 16.4, \; 13.9, \; 7.1 \]
Figure (svg): A vertical bar graph of the percentage of U.S. TikTok users in five age groups, separated bars: 10 to 19, 32.5; 20 to 29, 29.5; 30 to 39, 16.4; 40 to 49, 13.9; 50 and over, 7.1. The vertical axis runs from 0 to 35 percent.
Notice that the age groups do have an order, and the bars follow it. They are still drawn apart, because the groups are not equal-width intervals: the last one, 50 and over, is open-ended. A histogram with touching bars would need equal, closed intervals.
Check that the heights are proportional
Why: the tallest bar divided by the shortest should match the ratio of the drawn heights.
\[ \dfrac{32.5}{7.1} = 4.58 \]
The 10 to 19 bar is about four and a half times as tall as the 50 and over bar, as it should be. In context: about a third of U.S. TikTok users are teenagers, and the share falls with every older age group.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 72-73 — Example 2.5, Table 2.9, Figure 2.3
Worked example
Park City's population is split into children, working-age adults and retirees. Table 2.10 gives counts and rounded percentages, and the book asks for a bar graph of the proportions. Recompute the percentages from the counts rather than trusting the printed ones.
Total the population
Why: the three groups make up the whole town.
\[ 67059 + 152198 + 131662 = 350919 \]
Share of children
Why: count over the whole town.
\[ \dfrac{67059}{350919} = 0.1911 \approx 19.1\% \]
Share of working-age adults
Why: same denominator.
\[ \dfrac{152198}{350919} = 0.4337 \approx 43.4\% \]
Each share uses the same denominator, the whole town. Dividing children by adults, or by the retirees, would give a ratio between groups, which is a different and less useful number for a bar graph of proportions.
Share of retirees
Why: same denominator again.
\[ \dfrac{131662}{350919} = 0.3752 \approx 37.5\% \]
The book's 19, 43 and 38 percent are these values rounded to whole numbers, so they agree.
Figure (svg): A vertical bar graph of the Park City population in three age groups: children 19 percent, working-age adults 43 percent and retirees 38 percent, with a vertical axis from 0 to 50 percent.
Check that the shares fill the whole town
Why: the three groups do not overlap and nothing is missing, so the shares should add to 100.
\[ 19.1 + 43.4 + 37.5 = 100.0, \qquad 19 + 43 + 38 = 100 \]
In context: working-age adults are the largest group in Park City at 43 percent, but retirees, at about 38 percent, outnumber children by roughly two to one.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 73 — Try It 2.5, Table 2.10
Worked example
Table 2.11 gives the projected number of U.S. high school graduates in 2030 for four regions, with percentages. The book asks for a bar graph with region, a qualitative variable, on the horizontal axis and the percentage on the vertical axis.
| region | graduates | percent |
|---|---|---|
| Northeast | 517,720 | 16.1% |
| Midwest | 695,170 | 21.6% |
| South | 1,253,540 | 39.0% |
| West | 749,400 | 23.3% |
Total the graduates
Why: the four regions cover the country.
\[ 517720 + 695170 + 1253540 + 749400 = 3215830 \]
Recompute the Northeast and Midwest shares
Why: count over the national total.
\[ \dfrac{517720}{3215830} = 16.10\%, \qquad \dfrac{695170}{3215830} = 21.62\% \]
Recomputing from the counts is how you catch a misprinted percentage, and it costs four divisions. Here the counts are the primary data and the percentages are derived from them, so the counts win any disagreement.
Recompute the South and West shares
Why: same denominator.
\[ \dfrac{1253540}{3215830} = 38.98\%, \qquad \dfrac{749400}{3215830} = 23.30\% \]
Figure (svg): A vertical bar graph of projected 2030 high school graduates by U.S. region: Northeast 16.1 percent, Midwest 21.6, South 39.0 and West 23.3, with a vertical axis from 0 to 45 percent.
Compare the largest and smallest regions
Why: divide the two percentages.
\[ \dfrac{39.0}{16.1} = 2.42 \]
Check the shares against the whole
Why: the regions are separate and complete, so the printed percentages should total 100.
\[ 16.1 + 21.6 + 39.0 + 23.3 = 100.0 \]
Every printed value matches the recomputed one. In context: almost two in five of the projected 2030 graduates are in the South, about two and a half times the Northeast's share.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 73-74 — Example 2.6, Table 2.11, Figure 2.4
Trap
A report wants the South to look dominant, so it starts the vertical axis at 15 percent.
\[ 39.0 - 15 = 24.0 \]
\[ 16.1 - 15 = 1.1 \]
The drawn bars are 24 and 1.1 units tall, so the South looks about 22 times the Northeast.
Bar length must be proportional to the value, so measure from zero.
\[ \dfrac{39.0}{16.1} = 2.42 \]
\[ \dfrac{24.0}{1.1} = 21.8 \]
The true ratio is under two and a half. The cut axis multiplied the visual difference about nine times.
Every number on the graph can be correct and the picture still lies.
Picture it
Here is the trap drawn. Both graphs use the four regional percentages from Example 2.6, every value correct, every bar labelled honestly.
Figure (svg): Two bar graphs of the same four regional percentages. The left one's vertical axis starts at 0 and the South bar is about two and a half times the Northeast bar. The right one's axis starts at 15, so the Northeast bar is a stub and the South bar looks about twenty times as tall.
The only difference is the bottom of the vertical axis. On the left it starts at zero, and the bars are proportional to the percentages: the South is a little under two and a half Northeasts. On the right the axis starts at 15, so each bar shows only the part of its value above 15, and the Northeast's 16.1 percent becomes a stub.
Your eye compares bar lengths, not the numbers printed on them. That is why a bar graph's axis must start at zero while a line graph's need not: a line graph asks you to read positions and slopes, which survive a shifted axis, but a bar graph asks you to compare lengths, which do not.
When you meet a bar graph in the wild, look at the bottom of its vertical axis before you look at the bars. It is the fastest check on whether the picture can be trusted.
Tweak it
The slider makes the exaggeration a curve. The horizontal axis is where the vertical axis starts, from 0 up to 15. The vertical axis is how many times taller the South bar looks than the smaller bar. The slider sets the smaller bar's true value.
Parameter explorer
Keep the smaller value at 16.1, the Northeast. At what starting point does the South bar look three times as tall? Then raise the smaller value to 23.3, the West, and see how much less the cut axis distorts.
\[ \text{apparent ratio} = \dfrac{39.0 - \text{start}}{{s} - \text{start}} \]
\[ \dfrac{39 - b}{16.1 - b} = 3 \;\Rightarrow\; 39 - b = 48.3 - 3b \;\Rightarrow\; b = 4.65 \]
Starting the axis at only 4.65 is enough to make a difference of under two and a half look like three. The closer the starting point gets to the smaller bar's value, the faster the ratio explodes, because the smaller bar's drawn length shrinks towards nothing.
The curve is flat at the left for a reason: with the axis at zero, the apparent ratio is the true ratio. Any honest bar graph sits at that left edge.
Worked example
Park City has six voting districts. Table 2.12 gives each district's share of the registered voters and its share of the whole population, and the book asks for a bar graph of the registered voter shares. Drawing both shares as grouped bars answers a sharper question.
| district | 1 | 2 | 3 | 4 | 5 | 6 |
|---|---|---|---|---|---|---|
| registered voters (%) | 15.5 | 12.2 | 9.8 | 17.4 | 22.8 | 22.3 |
| whole population (%) | 19.4 | 15.6 | 9.0 | 18.5 | 20.7 | 16.8 |
Check each column adds to 100
Why: both rows are shares of one whole: all voters, and all residents.
\[ 15.5 + 12.2 + 9.8 + 17.4 + 22.8 + 22.3 = 100.0 \]
The same check for the population row
Why: a second complete split of the city.
\[ 19.4 + 15.6 + 9.0 + 18.5 + 20.7 + 16.8 = 100.0 \]
Figure (svg): A grouped bar graph for Park City's six voting districts. In each district a blue bar shows the share of registered voters and an orange bar the share of the whole population: district 1, 15.5 and 19.4; 2, 12.2 and 15.6; 3, 9.8 and 9.0; 4, 17.4 and 18.5; 5, 22.8 and 20.7; 6, 22.3 and 16.8.
Subtract to compare the two shares in each district
Why: voter share minus population share; positive means the district holds more of the voters than of the people.
\[ -3.9, \; -3.4, \; +0.8, \; -1.1, \; +2.1, \; +5.5 \]
The grouped bars make the same comparison visually. In every district your eye can measure the blue bar against its orange neighbour, which a graph of the registered voters alone could never show.
Find the extremes
Why: district 6 has the biggest surplus, district 1 the biggest shortfall.
\[ 22.3 - 16.8 = 5.5, \qquad 15.5 - 19.4 = -3.9 \]
Check the differences balance out
Why: both rows total 100, so the six differences must add to zero.
\[ -3.9 - 3.4 + 0.8 - 1.1 + 2.1 + 5.5 = 0 \]
They do. In context: district 6 holds 22.3 percent of Park City's registered voters but only 16.8 percent of its residents, while district 1 has fewer voters than its population share would suggest, a gap a campaign or a registration drive would want to see.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, pp. 74-75 — Try It 2.6, Table 2.12
Concept
The book allows bars in either direction. The choice is practical. Vertical bars suit a few short category names and read naturally for values that grow upward. Horizontal bars suit long category names, because each name gets a full line to itself instead of being squeezed or tilted under a narrow bar.
Figure (svg): A horizontal bar graph of projected graduates by region, sorted from largest to smallest: South 1,253,540; West 749,400; Midwest 695,170; Northeast 517,720. Region names sit at the left of each bar with the count at its right.
Order is the other free choice, because categories have no natural order. Sorting the bars from largest to smallest turns the graph into a ranking you can read from the top: the South first, the Northeast last. Chapter 1 called a bar graph sorted this way a Pareto chart.
This version also shows counts rather than percentages, which is equally legitimate. The bars have the same proportions either way, since every percentage is its count over the same total of 3,215,830.
\[ \dfrac{1253540}{517720} = 2.42 = \dfrac{39.0}{16.1} \]
When categories do have an order, such as the TikTok age groups, keep it. There the size order and the age order happen to agree, but for a table where they did not, sorting by size would hide how the shares change with age, which is usually the most interesting fact.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 72 — bars can be vertical or horizontal
Comparison
Fill in the gaps to pull the section together. Each row is a display; each column is a question you should be able to answer about it without looking anything up.
Comparison matrix
| display | kind of data | keeps every value? | what the picture claims |
|---|---|---|---|
| stemplot | a small set of numbers | yes | every stem written, so gaps are real |
| side-by-side stemplot | two sets, same units | yes | shared stems |
| line graph | ordered | no, frequencies only | the axis is ordered, so rises and falls mean something |
| bar graph | categories | no | separate categories; lengths from zero |
The keeps-every-value column is the stemplot's advantage and its limit. Only the stemplots keep the original data, and that is only practical for small data sets; the line and bar graphs give up individual values in exchange for handling any number of observations.
The last column is the one to remember when judging someone else's graph. Each display makes a claim by its form: empty stems claim a gap, joined points claim an order, separated bars claim distinct categories, and bar lengths claim proportion. A graph is misleading when its form claims something the data do not support.
Counterexample
The book recommends stemplots for small data sets. That condition is easy to state and easy to forget, so build a case where the stemplot fails.
Discussion prompt
Describe a data set for which a stemplot would be a poor display, say exactly what goes wrong, and name a better choice.
Two signs warn you off a stemplot: many more values than you could write in a few lines, and a range so wide that most stems would be empty. Either one breaks the promise that row length shows shape.
The rule of thumb in Section 2.2 is that a histogram suits data sets of a hundred values or more. Under that, and with a sensible range, the stemplot is the better picture, because it keeps every value.
OpenStax Introductory Statistics 2e, §2.1 Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar Graphs §2.1, p. 66 — a good choice when the data sets are small
Real world
Bar graphs are the most common statistical picture in news reports and advertising, and the truncated axis is the most common trick played with them. Practise the two-question check on a realistic case.
Discussion prompt
A news site shows unemployment in two cities as bars: City A at 4.2 percent and City B at 4.8 percent. City B's bar is drawn four times as tall as City A's. Find where the vertical axis must start for that to happen, and write the sentence an honest graph would support.
The habit to take away is to look at the vertical axis before the bars. If it does not start at zero, divide the real values yourself before believing the picture.
Line graphs of a rate over time often start their axis well above zero, and that is acceptable for a line, since you read its rises and falls rather than its lengths. The rule is specific to bars.
Section
Part 6
Pattern
Everything in this section reduces to a short chain of questions. Work through them in order whenever you have data to show.
Figure (svg): A decision flow chart. The first box asks what kind of data: categories or numbers. Categories lead to a bar graph with separated bars. Numbers lead to a second box asking whether you want every value kept and the set is small; yes leads to a stemplot, or a side-by-side stemplot for two sets on one spine. The other branch, a few distinct values with their frequencies, leads to a line graph of frequencies.
The last two rows are the ones most often skipped, and they are where the marks are. A graph that is never read has answered nothing, and a graph that was never checked may be showing a miscopied number.
Check
A stemplot of times in seconds has the key 12 bar 4 means 12.4 seconds. Decide what one row holds before reading the choices.
Check your understanding
The row 15 | 0 3 3 8 appears in the plot. Which values does it represent?
Answer: A
Why: With the key 12 bar 4 meaning 12.4, each leaf is a tenths digit, so the row holds four separate times: 15.0, 15.3, 15.3 and 15.8 seconds. The repeated 3 is two different observations that happen to be equal.
Two of the wrong answers are the misreadings from Part 1: dropping a tie, and reading a row as one number. The third is why every stemplot needs a key, since the digits alone cannot tell 15.3 from 153.
If you hesitated, rebuild one value: stem 15 and leaf 8, with the leaf in the tenths place, is 15.8. Rebuilding is always the test of a reading.
Check
A judgement about a gap, the way Part 2 made it. Look at the spacing, not just the size.
Check your understanding
Twelve waiting times in minutes are 8, 12, 15, 15, 17, 21, 22, 22, 24, 26, 29 and 47. Which statement does a stemplot of these data best support?
Answer: A
Why: The stems run 0 to 4, and stem 3 is empty. The gap from 29 to 47 is 18 minutes, while every other step between neighbours is at most 4 minutes, so 47 does not fit the pattern and may be an outlier.
The last distractor is worth a second look. It names the correct value, and an examiner would still mark the reasoning wrong, because a lone leaf does not make a gap; the empty stem does.
\[ 47 - 29 = 18, \qquad \text{other steps} \le 4 \]
Check
The last decision of the lesson, the one that comes before any drawing. Ask whether the horizontal axis has a natural order.
Check your understanding
A survey records the type of pet each of 120 households owns: dog, cat, fish, bird or none. Which display is appropriate for the counts?
Answer: A
Why: Pet type is a category with no natural order, so the counts belong in separate bars, and the bars must start at zero so that their lengths are proportional to the counts.
Two of the wrong options would produce a perfectly neat graph. That is the danger with displays: a wrong choice rarely looks wrong. It looks like a graph making a claim, and the claim is false.
If you chose the line graph, return to the blood-type figure, where the same four numbers drew two opposite trends.
Explain it to yourself
This section's deepest idea is that a graph's form makes claims about the data. Saying that in your own words is what lets you judge a graph you have never seen.
Discussion prompt
In three or four sentences, explain why a stemplot must include its empty stems, why a line graph may only join points over an ordered axis, and why a bar graph's bars must start at zero. Try to give one underlying reason that covers all three.
If your answer named the eye or the reader, it is on the right track. Graphs exist to be read quickly, so their rules are about making the quick reading the correct one.
If it listed three separate rules with no link, try the one-sentence version: the form of the picture must not claim more than the numbers do.
Exit ticket
One small data set, the whole lesson. Answer in full sentences, and name the evidence for each claim.
A coffee shop records the waiting times, in minutes, of twelve customers: 8, 12, 15, 15, 17, 21, 22, 22, 24, 26, 29 and 47.
Discussion prompt
Build the stemplot with a key, say where the times concentrate and whether any value may be an outlier, and say whether a line graph or a bar graph of these twelve times would be a sensible alternative.
The most common slip is leaving out stem 3. Without it, 47 sits directly under the twenties and looks like an ordinary value.
The second is calling 8 an outlier because it is lowest. The step from 8 to 12 is ordinary; only the empty stem shows a gap.
Recap
This section built three pictures and, more importantly, the rules that make each one honest. Each row of the table is one display, with the check that keeps it right and the mistake to watch for.
| display | build it | check | the trap |
|---|---|---|---|
| stemplot | leaf is the final digit; every stem; leaves in order; key | leaves add to n | dropping empty stems |
| reading a stemplot | concentration, spread, gaps | measure the gap | calling the largest value an outlier |
| side-by-side stemplot | one shared spine; smallest leaves nearest it | each side adds to its n | reversing the right side |
| line graph | ordered values across, frequencies up, joined | frequencies add to the sample | joining categories |
| bar graph | separated bars, any helpful order, from zero | percents add to 100 | a cut axis |
\[ \text{form of the picture} \;\Rightarrow\; \text{a claim about the data} \]
That line is the lesson in miniature. Empty stems claim gaps, joined points claim order, separated bars claim categories, and bar lengths claim proportion. Build every graph so that its claims are true, and read every graph by asking whether they are.
Section 2.2 picks up where the stemplot stops: histograms for large data sets, frequency polygons, which are line graphs of grouped data, and time series graphs.
OpenStax Introductory Business Statistics 2e, §2.1 Display Data §2.1 — the same displays applied to business data
Want this taught 1-on-1? Alexander tutors Statistics — $55/session, free consultation.