Side-by-Side Bar Charts for Two Groups

Comparing the Same Categories Across Two Groups

A bar chart can show the distribution of one categorical variable for one group. When you want to compare that distribution across two groups, a side-by-side bar chart places the bars for the groups next to each other for each response category. This arrangement makes it easier to compare the groups category by category.

In Misleading Axes and Scales in Bar Charts, you learned to check the scale and bar design before interpreting a display. For a side-by-side chart, also check that the categories line up across the groups and that a key makes the groups easy to distinguish. The bars should have equal widths, visible gaps between categories, and a vertical scale that starts at zero.

For example, a survey might record a preferred activity and the age group of each respondent. The activity is the response category being compared, and age group separates the data into groups. The horizontal axis lists the activity categories. Within each activity, one bar represents each age group; a legend identifies which bar belongs to which group.

The category order should be the same for both groups. If categories have a natural order, such as “never,” “sometimes,” and “often,” preserve it. Otherwise, choose a clear order and use it consistently. Keep the group colors or patterns consistent throughout the chart and explain them in a legend.

Choose Counts or Relative Frequencies

A side-by-side chart can display counts or relative frequencies. Counts answer questions about how many individuals in each group fall into each category. Relative frequencies show what fraction or percentage of each group falls into each category.

The choice matters when the groups have different numbers of individuals. A larger group may have larger counts in several categories simply because more individuals were surveyed. If the goal is to compare the groups’ distributions, use relative frequencies calculated separately within each group. This is the same denominator idea used in the earlier tutorials on conditional distributions and choosing the correct denominator.

When the chart uses relative frequencies, all bars use the same vertical scale, often from \(0\%\) to \(100\%\). Within each group, the relative frequencies across all response categories add to \(100\%\), allowing a direct comparison of the groups’ distributions. In a count chart, by contrast, each group’s category counts add to that group’s total.

A count chart is useful when the question concerns numbers of individuals or when group sizes are similar and the comparison of counts is meaningful. A relative-frequency chart is usually more informative for comparing how responses are distributed when group sizes differ. As in Comparing Conditional Percentages in Context, describe differences in percentages as percentage-point differences.

How to Build a Side-by-Side Bar Chart

Start with a frequency table that lists the same response categories for each group. Check that every individual is counted once in an appropriate category, and that each group’s category counts add to its group total. If you are comparing relative frequencies, divide each count by its own group total before drawing the bars.

The side-by-side arrangement keeps the comparison focused on a response category: look across the neighboring bars to compare the groups for that category. Then move to another category. Do not confuse this with comparing categories within just one group; both comparisons can be useful, but they answer different questions.

Worked Example: Build a Count Chart

Worked Example: Use Relative Frequencies When Group Sizes Differ

Worked Example: Read a Side-by-Side Chart in Context

Common Mistakes and What a Strong Answer Says

• Using counts to compare unequal group sizes without noting the limitation. Larger groups often have larger counts. If the question is about distributions, calculate each group’s relative frequencies and compare those instead.
• Using the grand total as the denominator for a group percentage. To find the percentage within one group, divide by that group’s total, not the total across both groups. State the group and category so the denominator is clear.
• Comparing bars from different categories as if they answer the same question. First compare the adjacent group bars within one category. Then, if useful, describe the most common category within each group.
• Leaving the group key or measurement unclear. Label the groups and say whether heights represent counts or percentages. Without those details, a reader may not know what a bar means.
• Calling a percentage difference a percent difference. Subtracting \(40\%\) and \(30\%\) gives a difference of 10 percentage points. A strong answer identifies it that way and names the groups and category.
• Claiming that a chart proves a cause. A side-by-side chart displays patterns in the data. As in the earlier tutorial on association between categorical variables, differences in distributions may show an association in the observed data, but the graph alone does not establish causation.

A strong comparison does more than say that one bar is taller. It names the category and both groups, gives the relevant counts or percentages, and describes the difference in context. If group sizes are unequal, it uses within-group relative frequencies when comparing distributions.