Guidelines
Suggest edit

Mirror bar small multiples for correlation comparison

For comparison of similarity between two quantitative series in paired bar charts, prefer mirrored small-multiple alignment to improve fidelity and mitigate cross-panel correspondence errors for viewers judging which pair is more correlated.

  • purpose:refine
  • basis:empirical
  • task:compare
  • chart:bar
  • structure:small-multiples
  • lever:layout-structure
  • operator:association
  • group-cardinality:binary

advice

Mirrored bar pairs for similarity

Mirror each pair of bar charts when viewers need to decide which pair is more similar or correlated. For example, center-align the two series and reverse one x-axis instead of keeping both charts in a standard adjacent left-to-right pair.

reason

Why mirrored bar small multiples work for this task

Correlation judgments depend on matching corresponding values across the pair, and mirror symmetry makes those matches easier to read.

Mechanism: Centered mirror alignment reduces cross-panel matching effort and supports faster, more precise comparison of overall similarity.

Evidence: In the bar-chart correlation experiment, mirrored small multiples let participants succeed at a lower target correlation than adjacent small multiples, and the paper reports no benefit for animation in this task (Ondov et al., 2019).

context

Use when all of these are true

  • User Goal: Decide which bar-chart pair is more similar or more correlated.
  • Task: Compare pairs of bar-chart series rather than single values.
  • Data: Two quantitative series per pair, with comparable means and standard deviations.
  • Chart Setting: A static small-multiple bar layout is being used for a two-way comparison.
  • Audience: Viewers inspect one paired comparison at a time.
  • Success Criterion: Readers can distinguish higher from lower correlation at smaller correlation differences.

exceptions

Do not use when any of these are true

Break it when: You must compare more than two datasets at once. Why: The paper notes that mirroring implies only two datasets and is unlikely to scale.

costs

Tradeoffs of mirrored bar pairs

Sacrifice: A conventional same-direction x-axis on both charts. Risk: The mirrored view is less scalable for many simultaneous comparisons. Mitigation: Use it for direct two-way correlation judgments only.

mistakes

Common failure mode

Mistake: Keep a same-direction adjacent pair when the task is to judge overall similarity between two bar-chart series. Why it fails: It gives up the center-aligned mirror structure that improved correlation comparison.

check

How to test this choice

Failure Sign: Readers need very high similarity before they can reliably pick the more correlated pair. Quick Check: Show mirrored and standard adjacent versions of the same bar-chart pairs, then ask readers which pair is more similar. Stronger Test: Reduce the correlation gap between the candidate pairs and keep the layout that still produces correct choices.

fix

What to change

  • Reverse the x-axis direction of one chart in each pair.
  • Center-align the two charts so corresponding bars face each other.
  • Keep the mirrored view limited to two-way comparisons.
  • If you must show many datasets at once, move back to a non-mirrored layout.

References

Ondov, B., Jardine, N., Elmqvist, N., & Franconeri, S. (2019). Face to Face: Evaluating Visual Comparison. IEEE Transactions on Visualization and Computer Graphics, 25(1), 861–871. https://doi.org/10.1109/TVCG.2018.2864884