Points of View: Unentangling complex plots

Journal name:
Nature Methods
Volume:
12,
Page:
591
Year published:
DOI:
doi:10.1038/nmeth.3451
Published online

Carefully designed subplots scaled to the data are often superior to a single complex overview plot.

At a glance

Figures

  1. Small multiples and progressive cropping helps to compare data traces across various y-axis ranges.
    Figure 1: Small multiples and progressive cropping helps to compare data traces across various y-axis ranges.

    (a) Categories and patterns can be difficult to distinguish when all the time-course response data are in a single plot. (b) Small-multiple plots isolate and untangle the categories but lose context as categories are separated. (c) Subtle scale annotations provide context while maintaining clarity.

  2. Make design choices that show trends in context.
    Figure 2: Make design choices that show trends in context.

    (a) In a single panel, categories with the widest ranges are often the most prominent. (b) Small multiples help navigation and simplify encodings. (c) Perception of differences is compromised across multiples for which minima, maxima or range vary. (d) Relative changes within categories are emphasized when panels are scaled within each category, but those between categories are difficult to judge. (e) Scaling to each category's maximum while using a global minimum contextualizes variation between categories. (f) Use an overview and scaled detail to contextualize, highlight and examine each category. Colored backgrounds emphasize differences in scale expansion. A vertical layout helps in identifying common changes in patterns.

References

  1. Shoresh, N. & Wong, B. Nat. Methods 9, 5 (2012).
  2. Pandey, A.V., Rall, K., Satterthwaite, M.L., Nov, O. & Bertini, E. in Proc. CHI Conf. Hum. Factors Computing Syst. 14691478 (ACM, 2015).
  3. Cleveland, W.S., McGill, M.E. & McGill, R. J. Am. Stat. Assoc. 83, 289300 (1988).
  4. Krzywinski, M. & Cairo, A. Nat. Methods 10, 687 (2013).

Download references

Author information

Affiliations

  1. Gregor McInerny is a Senior Research Fellow at the Department of Computer Science, University of Oxford.

  2. Martin Krzywinski is a staff scientist at Canada's Michael Smith Genome Sciences Centre.

Author details

Additional data