Making Data Visualization Work for You

Principles of Good Data Viz for the Busy Scientist



Julia Piaskowski

Feb 14, 2025

https://jpiaskowski.gitlab.io/talks/data-viz-guide/

Part 1: Using Data Visualization to Explore Your Data

EDA: Exploratory Data Analysis

  • Look at your data and make sure all is okay

  • This is (hopefully) a quick and easy process - use whatever software you feel most comfortable with

  • It doesn’t have to be pretty

  • Check every variable and covariate you plan to analyze

Distributions

Distributions

Cross Tabulations

Pairwise Relationships

Pairwise Relationships

Change Over Time

Part 2: Use Data Visualization to Make a Point

A Good Data Vizualization Will

  • Show the data - literally, let’s see some results!
  • Induce the viewer to focus on substance, not the design
  • Not distort the data
  • Present many numbers in a small space
  • Make large data sets coherent
  • Encourage viewers to compare different pieces of data
  • Reveal data at several levels, from a broad overview to fine detail
  • Serve a clear purpose
  • Be closely integrated with statistical methods and/or descriptions of a data set (reference them in your manuscript)

Image by Florence Nightingale

Soil Texture Triangle

Slave Voyages

Use Data Viz to Convey a Clear Point

  • Think about the message(s) you want the plot to convey
  • Draw your desired plot first on paper

Avoid Distorting Your Data

Present a Critical Mass of Data

Make Your Data Viz Self Explanatory

Don’t make readers struggle.

New York Times

Make Your Data Viz Self Explanatory

Simplify the plot:

- Remove unneeded decoration: shadows, non-informative 3D, gratuitous color

_ Consider if a simple background with minimal or no gridlines will work

Label your plot:

- All axes

- Title and a legend to add extra information

- Include just enough information to understand the plot


Plots Need to Be Interpretable by Themselves

Avoid Unneeded 3D

Use Informative 3D

Avoid Confusing or Unneeded Color

Avoid Excessive Numbers of Colors

Use Color to Inform

Make Your Data Viz Accessible

  1. Make font sizes, plotting symbols and line widths large enough for people to read
    • At least equivalent to 11-12 points in journals, larger for posters/presentations
  2. Use colors that work for all people
    • ‘Viridis’ is a color-blind friendly set of palettes
  3. Ensure sufficient contrast between colors or at least the lightest color and your background

How to Check Accessibility

Make Your Plot Pop

Use something other than defaults.

Use Points as an Alternative to Bars


Use An Uncommon Plot

the Ridge Plot

Use an Uncommon Plot: the Rain Plot

Be Consistent with Colors

Avoid Repetitiveness


Avoid having the same chart type for everything (e.g. a bar plot)

Final Thoughts

  • Use plotting to explore and understand your data
  • Think through the information and message you want communicated with a plot. Don't do a plot just because everybody else presents the same plot.
  • There are fewer rules than you might think: error bars don’t have to be the standard error - they can show the variance. It depends on what you want to say with that plot!
  • Be creative: browse plotting libraries and other journals for ideas

Additional Resources