Use dot plots instead of bar graphs for average comparisons with unequal group sizes

Use a dot plot instead of a filled bar graph when readers must decide which group has the higher average and the groups contain different numbers of observations. For example, show each value as a dot on a shared y-scale rather than as a set of filled bars, because bar area can stand in for the sum instead of the average.

Filled bars carry both value and area, so a group with more bars can look larger in total even when that does not match the average. Dots remove the filled-area cue and make the comparison depend more on overall position.

Mechanism: Unequal group sizes break the correlation between total filled area and average value. In bars, that mismatch harms mean judgments; dots reduce the area cue and support comparing average position more directly.

Evidence: In unequal set-size comparisons, dot graphs produced better discrimination than normal bar graphs, while normal bar graphs performed like misaligned length-only bars; performance also dropped sharply when unequal counts made summed area a poor proxy for average (Yuan et al., 2019).

Notes: The experiments did not show a reliable dot-plot advantage when the compared groups had equal numbers of items.

• User Goal: Decide which of two groups has the higher average.
• Task: Compare group means.
• Data: Quantitative observations shown as multiple raw values per group, with unequal numbers of observations across groups.
• Chart Setting: Static displays where each group’s raw values are visible.
• Success Criterion: More accurate average comparisons when group sizes differ.

• Break it when: The task is a single-value comparison rather than a group-mean comparison. Why: In 1vs1 comparisons, aligned bars were as precise as dots.
• Break it when: The compared groups have the same number of observations and unequal-count bias is not present. Why: The experiments did not show a reliable dot-plot advantage in that condition.

Sacrifice: You replace aggregate-looking rectangles with individual points. Risk: Unequal group size can still bias judgments in dot plots, because the more numerous group can influence the decision. Mitigation: If the mean itself is the key message, represent the mean explicitly in addition to the dots.

Mistake: Using filled bars for unequal-size raw groups and expecting readers to extract the mean from bar tops alone. Why it fails: Readers can use total bar area as a proxy, and that breaks down when group sizes differ.

Failure Sign: The group with more marks also looks like the larger total, even when its average is close to or below the other group. Quick Check: Build a test case where the higher-average group has fewer observations and see whether the bar version becomes much harder to judge than a dot version. Stronger Test: A/B test the same unequal-size comparison as a bar graph and as a dot plot, and compare accuracy on which group average is higher.

• Replace each bar with a dot placed on the same value scale.
• Keep the two groups separate, but remove filled bar area as the dominant cue.
• Add an explicit mean representation if the chart’s question is about the average.
• If the graphic must stay bar-based, avoid using it as the sole display for judging unequal-size group averages.