Why Group Size Changes What Counts Tell You
In Mosaic Plots and Bar Widths, you saw that group size and within-group percentages convey different information. A large group can have more individuals in a category simply because it contains more individuals overall. If the question is which group has a greater share in that category, raw counts do not answer it: compare percentages calculated within each group.
A percent graph puts groups on a common scale. Each group’s categories are represented as percentages of that group, so its total is \(100\%\), regardless of how many individuals are in it. This makes the relative distribution easier to compare when the group sizes differ. Counts are still useful when the question is how many individuals fall in a category; percentages answer how common that category is within a group.
To calculate a group-specific percentage, divide a category’s count by that group’s total and multiply by \(100\%\). This uses the same within-group denominator principle from earlier tutorials on conditional percentages. In a two-way table, each percentage in the response distribution for a group uses that group’s row or column total, depending on how the table is arranged.
A percent graph can show one response category or the full distribution of response categories. A side-by-side percent bar chart places corresponding category bars next to one another, with heights measured on a common percentage scale. A segmented bar chart divides each group’s bar into response categories, with each whole bar representing \(100\%\). As covered in earlier tutorials, segmented bars are especially useful for comparing entire conditional distributions.
In either graph, the bar heights or segment sizes show percentages, not raw counts. The bars may be equal in width even when the group totals are different. That is intentional: the display focuses on the percentage pattern rather than the number of people represented. Include clear labels so readers know which groups and response categories are being compared.
Worked Example: More Counted Responses, Lower Percentage
Worked Example: More Counted Responses, Lower Percentage
A fictional job-skills program records whether participants completed an optional practice module. The afternoon group is larger than the morning group. Which group had the higher completion percentage, and what would the raw counts suggest?
| Session group | Completed | Did not complete | Total |
|---|---|---|---|
| Morning | 28 | 12 | 40 |
| Afternoon | 60 | 60 | 120 |
State. The question asks which session group had the higher share of participants completing the module. The relevant comparison is the completion percentage within each session, not the completion count alone.
Plan. Divide each completion count by its own session total. This puts both groups on a common percentage scale, even though one group has three times as many participants.
Do. The morning completion percentage is \(28/40=0.70=70\%\). The afternoon completion percentage is \(60/120=0.50=50\%\). The morning percentage is \(70\%-50\%=20\) percentage points higher.
Interpret. In this fictional program, 70% of morning participants completed the module, compared with 50% of afternoon participants, a difference of 20 percentage points. A side-by-side percent bar chart would show the morning completion bar reaching 70% and the afternoon bar reaching 50%.
Conclude. The raw counts alone point in the opposite direction: 60 afternoon participants completed the module, compared with 28 in the morning. But the afternoon group had three times the number of participants. Counts answer which session had more completions; percentages answer which session had the greater completion share.
Use the Same Scale for Every Group
A percent graph is useful only if each percentage uses the correct group total. Calculate the percentages within each group, then put all the groups on a common axis from \(0\%\) to \(100\%\). For a segmented bar chart, the response-category segments within each group must add to \(100\%\). For side-by-side percent bars, each category is compared using its percentage within each group.
The common scale makes comparisons visible without letting a large group take up more space or produce taller bars merely because it contains more people. A percent graph does not erase group-size information for every purpose; rather, it intentionally shifts attention to the distribution. If the group totals matter to the reader, report them alongside the graph or in a table.
Worked Example: Compare Full Distributions With Percent Bars
Worked Example: Compare Full Distributions With Percent Bars
A fictional community survey asks residents how often they bring a reusable cup to a local café. The sample includes 30 residents from one neighborhood and 90 from another. Calculate the within-neighborhood percentages and describe what a segmented percent bar chart would show.
| Neighborhood | Always | Sometimes | Never | Total |
|---|---|---|---|---|
| Riverside | 12 | 12 | 6 | 30 |
| Hillview | 18 | 45 | 27 | 90 |
State. The response variable is how often residents bring a reusable cup. The groups are the two neighborhoods. The goal is to compare each neighborhood’s response distribution, so calculate percentages using its own total.
Plan. For each response category, divide its count by the total for that neighborhood and convert to a percentage. Then check that the percentages in each neighborhood add to \(100\%\).
Do. For Riverside, the percentages are \(12/30=40\%\) for always, \(12/30=40\%\) for sometimes, and \(6/30=20\%\) for never. They sum to \(40\%+40\%+20\%=100\%\).
For Hillview, the percentages are \(18/90=20\%\) for always, \(45/90=50\%\) for sometimes, and \(27/90=30\%\) for never. They sum to \(20\%+50\%+30\%=100\%\).
| Neighborhood | Always | Sometimes | Never | Total percentage |
|---|---|---|---|---|
| Riverside | 40% | 40% | 20% | 100% |
| Hillview | 20% | 50% | 30% | 100% |
Interpret. In Riverside, 40% of surveyed residents always bring a reusable cup, compared with 20% in Hillview, a difference of 20 percentage points. In Hillview, 50% sometimes bring one, compared with 40% in Riverside. The never percentages are 30% in Hillview and 20% in Riverside.
Conclude. A segmented percent bar chart would give both neighborhoods bars of equal height, each representing \(100\%\), but the segment sizes would differ. It would show that the response distributions are not identical: Riverside has a larger always percentage, while Hillview has larger sometimes and never percentages. The graph compares these distributions, not the number of residents surveyed.
When Counts and Percentages Tell the Same Story
Different group sizes do not guarantee that counts and percentages will point in opposite directions. Sometimes the distributions are similar, and a larger group has both more individuals in a category and the same percentage in that category. A percent graph helps reveal this too: groups with identical percentages have matching bar heights or segment sizes, even if their counts differ.
This is a useful check against overinterpreting raw counts. Before claiming that one group is more likely to have a particular response, ask what fraction of each group has that response. If the percentages are equal, a larger count in one group may simply reflect that the group itself is larger. As in earlier tutorials on comparing conditional percentages, describe the observed difference in percentage points when comparing two percentages.
Worked Example: Different Counts, Matching Percentages
Worked Example: Different Counts, Matching Percentages
A fictional school reviews whether students who attended a study workshop felt that their quiz preparation improved. The two workshop sessions had different enrollment totals. Compare the counts and percentages reporting improvement.
| Workshop session | Preparation improved | Did not report improvement | Total |
|---|---|---|---|
| Early session | 15 | 10 | 25 |
| Late session | 45 | 30 | 75 |
State. Compare the percentage reporting improved preparation in each session. The session totals are unequal, so compute each percentage within its own session.
Plan. Divide the improvement count by the session total. Also check the remaining category to verify the complete two-category distributions.
Do. In the early session, \(15/25=60\%\) reported improvement and \(10/25=40\%\) did not. In the late session, \(45/75=60\%\) reported improvement and \(30/75=40\%\) did not.
Interpret. In both workshop sessions, 60% of surveyed students reported improved preparation. The late session had 45 such students, compared with 15 in the early session, but it also had three times as many students overall.
Conclude. A percent graph would show matching response distributions for the two sessions: 60% reporting improvement and 40% not reporting improvement in each. The larger late-session count does not mean its students had a higher improvement percentage. These survey results describe the students in the data; by themselves, they do not establish that attending a workshop caused improved preparation.
Common Mistakes and AP Exam Tips
- Comparing raw counts when the question asks which group has the higher share. Divide by each group’s own total before comparing. A group with more individuals can have a larger count but a smaller percentage.
- Using the grand total as the denominator. For a within-group percentage, use the total for the specified group. Dividing by the grand total gives a joint percentage for the whole data set, which answers a different question.
- Mixing counts and percentages in one comparison. Do not say one group has a “higher rate” merely because its category count is greater. Calculate and state the percentages if the question concerns the share of each group.
- Forgetting to check the distributions. When the categories cover all possible responses, the percentages within each group should add to \(100\%\), aside from rounding. A total far from \(100\%\) often signals an incorrect denominator or arithmetic error.
- Calling a difference in percentages a percent increase. Subtract the percentages to report a percentage-point difference. For example, \(70\%-50\%=20\) percentage points.
- Leaving out the groups, response, or denominator context. A full-credit comparison names both groups and the response category, reports the percentages, and gives the direction of the difference in context.
A strong AP-style sentence might say: “In this survey, 70% of morning participants completed the practice module, compared with 50% of afternoon participants, a difference of 20 percentage points. Although more afternoon participants completed it by count, the morning group had the higher completion percentage.” This makes clear what each comparison describes.
Check Your Understanding
Use the group total as the denominator when a question asks for a within-group percentage.
- In a sample of 40 cyclists, 24 wear a helmet on every ride. In a sample of 100 runners, 50 wear a hat on every run. Which group has the greater percentage in the stated category, and by how many percentage points?
- Two groups have 18 and 54 individuals who chose the same response. What additional information is needed to compare the response percentages?
- A group has 16 “yes” responses and 24 “no” responses. Find both percentages and check their total.
- What does a segmented percent bar chart show about group totals, and what does it show about response distributions?
- A group has a greater count in a category but a lower within-group percentage than another group. Explain how both statements can be true.