Guidelines
Suggest edit

Prefer a superimposed log-scaled shared view for multivariate time-series comparison

For comparison tasks on ordered time, prefer a superimposed log-scaled shared view on multivariate line charts to improve comparison speed and mitigate slow cross-row reading for readers comparing several series.

  • purpose:select
  • basis:empirical
  • task:compare
  • time:ordered-time
  • structure:single-view:use
  • structure:small-multiples:avoid
  • measure:multi
  • quality:readability:use
  • lever:layout-structure

advice

Use a shared comparison view

Use a superimposed log-scaled shared view when readers need to compare multiple time series quickly. For example, replace a row-separated linear line-plot layout with one shared line plot on a log-scaled y-axis, and keep color hue to distinguish the overlaid series.

reason

Why the shared log view is faster

A shared view keeps the lines close together, so readers can compare them directly instead of scanning across separate rows. In this contrast, the improvement is about speed, not a measured gain in correctness.

Mechanism: Superimposition reduces cross-row comparison work, and the tested shared log view supported faster answers on cross-series comparison tasks.

Evidence: In the collated extraction, the superimposed log-scaled line plot ranked above the row-separated linear line plot for correlate, aggregate, and overall time, with significant pairwise differences, while the corresponding accuracy contrasts showed no significant pairwise difference (Zeng & Battle, 2023; Aigner et al., 2011).

Notes: The supported finding is the tested whole-view contrast, not an isolated claim about log scaling alone.

context

Use when comparing multiple time series

  • User Goal: Compare several time series quickly.
  • Task: Judge cross-series correlation or aggregate behavior.
  • Data: Multiple ordered-time series shown together.
  • Chart Setting: A line-chart choice between a row-separated linear layout and a superimposed log-scaled shared view.
  • Audience: Readers performing visual comparison across several series.
  • Success Criterion: Shorter task completion time.

exceptions

Do not use when accuracy is the only goal

Break it when: The main success criterion is improved correctness rather than faster completion. Why: The tested contrast showed a time advantage for the superimposed log view, but not a significant accuracy advantage.

costs

Tradeoffs of the shared log view

Sacrifice: You give up the row-separated linear arrangement. Risk: The switch can be overclaimed as an accuracy improvement even though the measured difference was in time. Mitigation: Use this contrast when faster comparison is the target outcome and review it with the same task type you expect readers to perform.

mistakes

Common layout failure

Mistake: Keep multivariate time series in separate rows on a linear scale for correlation or aggregate comparison. Why it fails: The tested row-separated linear view took longer than the superimposed log-scaled shared view on the same task types.

check

Compare the two layouts directly

Failure Sign: Readers are slow when answering cross-series correlation or aggregate questions. Quick Check: Build two versions of the same line chart: one row-separated with a linear y-axis and one superimposed with a log y-axis, then compare answer times on the same questions. Stronger Test: Time a short set of correlation and aggregation tasks on both versions and keep the version with consistently shorter completion times.

fix

Change the layout and scale

  • Remove the row separation and place the series in one shared line plot.
  • Switch the shared y-axis from linear to log.
  • Keep color hue to differentiate the overlaid lines after superimposition.
  • Re-test the revised chart against the row-separated linear version on the same comparison tasks.

References

Aigner, W., Kainz, C., Ma, R., & Miksch, S. (2011). Bertin was Right: An Empirical Evaluation of Indexing to Compare Multivariate Time-Series Data Using Line Plots. Computer Graphics Forum, 30(1), 215–228. https://doi.org/10.1111/j.1467-8659.2010.01845.x
Zeng, Z., & Battle, L. (2023). A Review and Collation of Graphical Perception Knowledge for Visualization Recommendation. Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, 1–16. https://doi.org/10.1145/3544548.3581349