Statistics › Section B · Descriptive statistics › Topic B.1
Stemplots, line graphs, and bar graphs
Numbers in a list hide their own patterns; the right graph reveals them. This topic covers three displays you can build by hand - stemplots that keep every value, line graphs that trace frequency, and bar graphs for categories - and how to spot the graphs that mislead.
Choosing a display
The right display depends on what kind of data you have - a point recalled from earlier topics - and on what you want to see: individual values, category sizes, or a pattern across values. This topic covers three displays for small data sets and categorical data: the stemplot, which keeps every value; the line graph, which shows how frequency rises and falls across a few ordered values; and the bar graph, the standard picture for category counts. Later topics add histograms and box plots for larger quantitative data sets, where grouping and summarising become necessary.
Stemplots: every value, kept
A stem-and-leaf plot, or stemplot, displays quantitative data while preserving every original value. Split each value into a stem (the leading digit or digits) and a leaf (the final digit): 24 becomes stem 2, leaf 4. Write the stems in a vertical column, then record each value's leaf beside its stem. For two-digit data the stems are the tens; for three-digit data, the hundreds and tens. Because the leaves are the actual digits, a stemplot is a graph and a data list at once - you see the shape and can still recover every observation.
Reading a stemplot
Once built, a stemplot shows three things at a glance: where the values cluster (the busiest stems), how spread they are (the range of stems used), and outliers - observations that sit far apart from the rest of the data. An outlier may be a mistake, like writing 50 for 500, or a genuine extreme that needs its own explanation; either way it should be investigated, not silently deleted. Leaves within a stem are usually sorted so the plot is easy to read, and every stemplot must carry a key, because 2|4 could mean 24, 240 or 2.4 depending on the units.
Side-by-side stemplots
Stemplots shine at comparisons. A side-by-side stemplot puts two data sets on one picture: the shared stems run down the middle, with one data set's leaves listed to the left and the other's to the right. Comparing where each side is dense shows at a glance which group tends higher, whether the two spreads differ, and whether either group has its own outlier. The display works best when both data sets share the same units and similar ranges, so their stems line up naturally.
Worked example: building a stemplot
Fifteen students report minutes spent reading yesterday: 8, 12, 15, 15, 17, 21, 22, 24, 24, 28, 32, 35, 38, 41, 78. 1. Split each value: 8 is stem 0, leaf 8; 24 is stem 2, leaf 4. 2. The stems run from 0 to 7, and stems 5 and 6 stay in the column with no leaves, so the gap shows. 3. The sorted leaves: 0|8; 1|2 5 5 7; 2|1 2 4 4 8; 3|2 5 8; 4|1; 5|; 6|; 7|8. 4. Key: 2|4 means 24 minutes. 5. Check: \(1+4+5+3+1+1 = 15\) leaves, matching the 15 students. Most values lie between 12 and 41 minutes, and 78 stands far apart: an outlier to investigate.
Line graphs: frequency across values
For a discrete quantitative variable with few values, a line graph plots each value against its frequency and connects the points with line segments: value on the horizontal axis, frequency on the vertical. The shape of the line shows how frequency rises and falls. Suppose 25 students report their number of siblings: 0 siblings for 4 students, 1 for 9, 2 for 7, 3 for 4, and 4 for 1 - a total of 25. The line climbs to a peak at 1 sibling and falls away after it, a pattern easier to see connected than in a bare table. Line graphs suit few ordered values; with many values the segments blur into noise.
Bar graphs for categories
The bar graph is the standard display for categorical data: one bar per category, its height - or length, if the bars run horizontally - showing the category's count or proportion. The gaps between bars matter: they signal that the categories are distinct groups, not points on a continuous scale. Horizontal bars help when category names are long. When two groups of different sizes are compared, use proportions rather than raw counts so the comparison is fair - the same lesson as relative frequencies. Bars can also sit side by side or stack, when parts make up a whole.
Worked example: a bar graph of transport
A class of 40 students travels to school as follows: walk 14, car 10, bus 8, bike 5, other 3. 1. Check the total: \(14+10+8+5+3 = 40\). 2. Draw one bar per category, with gaps, heights equal to the counts. 3. Convert to proportions: \(14/40 = 0.35\), \(10/40 = 0.25\), \(8/40 = 0.20\), \(5/40 = 0.125\), \(3/40 = 0.075\). 4. Check: the proportions total \(0.35+0.25+0.20+0.125+0.075 = 1\). Walking is the largest single category at 35%. Note what the graph cannot say: whether walking is growing or shrinking - that would need data collected over time, not a category snapshot.
Misleading graphs
Graphs persuade, which is exactly why they get manipulated. Watch for a vertical axis that starts far above zero, exaggerating small differences; unequal interval widths, which make bar areas incomparable; three-dimensional and decorative effects, which distort the sizes your eye compares; and cherry-picked time windows that hide the fuller pattern. The defence is a habit: before trusting any graph, read the axes, the labels and the units - what is plotted, on what scale, starting where? Once the structure is clear, judge the picture.
Common mistakes to avoid
First, building a stemplot without a key: 2|4 is ambiguous until the reader knows it means 24. Second, forgetting to sort the leaves, which hides the plot's own shape. Third, giving bar graph bars no gaps, which wrongly suggests a continuous scale. Fourth, comparing raw counts between groups of different sizes instead of proportions. Fifth, trusting a striking graph before reading its axes - a truncated axis can turn a two-point difference into a visually towering gap. Sixth, deleting an outlier without investigating what it is.
Key takeaways
Match the display to the data. Stemplots keep every value and reveal clusters, spread and outliers; side-by-side stemplots compare two groups back to back; both need a key. Line graphs connect value-to-frequency points for a few ordered values. Bar graphs picture category counts or proportions, with gaps marking distinct categories, and proportions are the fair basis for comparing differently sized groups. In every case, read the axes before the picture: truncated axes, unequal intervals, decoration and cherry-picked ranges are the standard tricks of misleading graphs.
Flashcards
Flashcard
Stems are the leading digits, leaves the final digits, of every value
Test your understanding
Question 1 of 5
Which display keeps every individual data value, so nothing is lost?
Question 2 of 5
A stemplot of commute times has the key 1|6 = 16 minutes. Stem 4 carries leaves 0, 3 and 9. Those leaves stand for:
Question 3 of 5
Why do bar graphs have gaps between the bars?
Question 4 of 5
In a bar graph of how 60 shoppers paid, the card bar has height 21. What proportion of shoppers paid by card?
Question 5 of 5
A bar graph's vertical axis starts at 90% rather than 0. What is the danger?
