From a Categorical Graph to a Justified Claim
A categorical graph can show which response is most common or which group has the larger proportion in a response category. A strong answer does more than point to the tallest bar or say what “looks bigger.” It names the comparison, cites the graph values that support it, and explains what those values mean in context.
In Graphing Categorical Data With Technology, you learned to check that a graph displays the intended counts or conditional percentages. That distinction matters when making a claim. Counts tell how many individuals are in a category; within-group proportions tell what share of each group is in that category. If the groups have different sizes, their counts alone may not answer which group has the higher proportion.
A useful response usually has three parts: state the answer to the question, cite the relevant values for each group, and describe the comparison in context. When useful, report the difference in percentage points. As in Comparing Conditional Percentages in Context, subtract the two percentages to find that difference.
A Reliable Way to Read and Compare
Before writing, identify the variable whose categories appear in the graph and the groups being compared. Then identify the particular response category named in the question. For a segmented bar chart, each full bar represents one group, and each segment represents a response category’s share of that group. For a side-by-side bar chart, check whether the bars show counts or percentages before comparing their heights.
Read the values from labels when they are provided. Otherwise, use the axis scale to estimate them. If values are estimated from a graph, describe them as approximate rather than reporting unnecessary decimal places. If the graph gives percentages, report percentages; if it gives counts, do not call those counts percentages.
Name the outcome category and the groups. For example: Which group has the larger percentage choosing a particular response?
Read the value for that outcome in each group. Confirm whether the graph displays counts, proportions, or percentages.
For a claim about which group has a greater share, compare within-group proportions or percentages. Do not compare one group’s count with another group’s percentage.
Name the group with the higher proportion, report both values, and state their difference if that helps explain the comparison.
A graph describes the data shown; it does not automatically support a claim about a larger population or explain why a difference exists. Unless the question gives a sound basis for broader conclusions, keep the claim about the individuals or groups represented in the display. Do not turn a descriptive comparison into a claim that one group caused an outcome or that the difference is statistically significant.
Worked Example: Comparing Segments in a 100% Stacked Chart
Worked Example: Comparing Segments in a 100% Stacked Chart
A fictional school survey records whether students in two after-school programs prefer meeting indoors or outdoors. A 100% stacked chart shows that 64% of students in the science program prefer indoors, compared with 46% of students in the arts program. Which group has the larger proportion preferring indoor meetings, and what evidence supports the claim?
Identify the comparison. The response category is “prefer indoors,” and the groups are students in the science program and students in the arts program. Since each bar represents 100% of one program, the segment percentages are within-group percentages and can be compared directly.
Compare the values. The science-program percentage is 64%; the arts-program percentage is 46%. The difference is:
State a justified claim. Among the students represented in the survey, the science program has the larger proportion preferring indoor meetings: 64%, compared with 46% in the arts program. The science-program percentage is 18 percentage points higher.
This is stronger than saying “the science bar is bigger.” It identifies the response, names both groups, reports the values, and gives the comparison in context. Because the chart displays percentages, there is no need to reconstruct the groups’ counts to answer this question.
Worked Example: Why Counts Alone Can Give the Wrong Comparison
Worked Example: Why Counts Alone Can Give the Wrong Comparison
A fictional recreation survey asks participants whether they use a walking trail at least once a week. A side-by-side bar chart displays counts: 42 weekly users among 60 morning-session participants and 54 weekly users among 90 evening-session participants. The question asks which session has the larger proportion of weekly trail users.
Check what the graph displays. The bars show counts, but the question asks about proportions within each session. The group totals differ, so comparing 42 with 54 would answer which group has more surveyed weekly users, not which group has the larger share of weekly users.
Calculate the within-group percentages. Divide the weekly-user count by the total number surveyed in that session:
Compare and conclude. The morning session has the larger observed proportion of weekly trail users. In the survey, 70% of morning-session participants used the trail weekly, compared with 60% of evening-session participants, a difference of 10 percentage points.
The evening session has the larger count of weekly users—54 compared with 42—because more participants were surveyed in that session. But the proportion is larger in the morning session. This is why a claim about which group has a greater share needs a denominator for each group, as emphasized in Choosing the Correct Denominator for a Percentage.
Worked Example: Estimating Values From a Graph
Worked Example: Estimating Values From a Graph
A fictional community survey compares how residents in two areas usually travel to a local market. The graph has a vertical axis labeled “Percent within area,” marked every 10 percentage points. The bicycle segment appears to cover approximately 30% to 55% of the Northside bar and approximately 15% to 35% of the Southside bar. Which area appears to have a larger proportion of residents who usually bicycle to the market?
Read the segment heights. The bicycle segment is between the 30% and 55% cumulative boundaries for Northside. Its size is the upper boundary minus the lower boundary:
For Southside, the segment extends from approximately 15% to 35%, so its size is approximately:
Make a suitably qualified claim. The graph suggests that the Northside sample has a slightly larger proportion of residents who usually bicycle to the market: about 25%, compared with about 20% in Southside. The estimated difference is about 5 percentage points.
The values here are approximate because they were read from a scale rather than given as labels. Reporting them as exact values such as 25.0% and 20.0% would suggest more precision than the graph provides. The comparison is still justified as long as it is clear that the values are estimates. This segment-reading method is useful when a stacked graph labels cumulative boundaries rather than labeling every segment’s size, as discussed in Reading a Segmented Bar Chart.
Common Mistakes and Stronger Answers
- Comparing counts when the question asks for proportions. Counts depend on group size. Give the within-group percentages when the question asks which group has a greater share or how common a response is.
- Using the wrong denominator. A percentage for a group should describe the response count divided by that group’s total. Do not use the grand total unless the question asks for a share of everyone represented in the graph.
- Naming a group without citing evidence. “The morning group is higher” is incomplete if the values are available. Include the response category and relevant percentages for both groups.
- Reporting a difference without units. Subtracting two percentages gives a difference in percentage points. Say “10 percentage points higher,” not “10 percent higher,” unless calculating a relative percent change is specifically requested.
- Giving more precision than the display supports. For values estimated from a graph, use “about” or “approximately.” Do not turn a rough reading into a precise decimal.
- Overstating what the graph proves. A graph can show which group has the larger percentage in the displayed data. By itself, it does not show that the difference is statistically significant, that it applies to all people in a broader population, or that group membership caused the outcome.
For a full-credit descriptive answer, make the context explicit. For example: “Among the surveyed participants, 70% of the morning-session group used the trail weekly, compared with 60% of the evening-session group; the morning-session percentage is 10 percentage points higher.” This directly answers the question and supports the claim with comparable values. If the graph provides counts instead of percentages, calculate or read the within-group proportions before making this kind of claim.
Check Your Understanding
Use the ideas in this tutorial to explain what claim the graph values support.
- A segmented bar chart shows that 58% of Group A and 43% of Group B chose “reuse.” Write a justified comparison in context, including the percentage-point difference.
- A count graph shows 36 “yes” responses out of 48 people in one group and 45 “yes” responses out of 75 people in another. Which group has the larger count, and which has the larger proportion? Show the calculations.
- A graph’s vertical axis is labeled “Percent within club.” What does a 40% response segment represent for a club?
- A stacked bar segment runs from approximately 20% to 65% on the graph’s scale. Estimate the segment’s percentage and explain how you found it.
- A student writes, “Group A’s percentage is 8 percent higher than Group B’s,” after subtracting 52% from 44%. What wording should the student use to describe the subtraction accurately?