Data Visualisation is Very Important and you **should** care about it: Part 3

Part 3: What's the solution? For line, bar, pie charts, etc.

I've outlined the problems (see Part 1 & 2), so here come the answers! How can you communicate data in clear, concise, universally understood, aesthetically pleasing ways?

Here are a few simple rules to follow.

• Choose the right chart for your data
  • Use great sites like https://datavizproject.com/ which let you select your purpose (or even data type!) and show you the best plots for that data
• Don't use violin plots. Dr Angela Collier was correct when she said "They are hard to read and bad".
• Don't use pie charts unless you want to make a joke or something. Just use a bar chart.
  • Especially don't use 3D pie charts
• Don't use clashing colors that look bad together
  • E.g. neon green and magenta
  • Or just colors that clash or are otherwise suggestive/inappropriate
  • Unless you're using it for particular effect, as in Simon Scarr's impeccable visualisation of the toll of the Iraq war for example.
• Avoid low-contrast color combinations
  • E.g. yellow text on a white background is hard to read
  • E.g. cyan markers and sky-blue markers are hard to distinguish
• Keep colorblindness/color vision deficiency requirements in mind
  • Mainly don't mix red and green without some other signifier to differentiate the two
• Don't include anything that is not needed
  • Space is often a commodity; don't waste it
  • People may assume something in the plot is relevant when it is not and be confused unnecessarily.
• Use non-color-specific signifiers where possible
  • Instead use other visual differential devices
  • E.g. when you have two datasets on a scatter plot use visually distinct markers for the two
  • E.g. make use of different linestyles in line plots
  • E.g. use patterns in your bar charts
• Keep the fonts large and legible and common
  • Keep things simple and boring because those things are easy
• Make sure that every aspect of your plot serves some sort of purpose
  • E.g. If your bar chart has colors those colors should mean something
• Avoid copyrighted material
• Don't use pie charts

The greatest crime you can commit in DataViz, however, is not making a bad plot, it's making a misleading plot. Data that is hard to read is not good, but data that misleads or lies to the audience is actively terrible.

I have, somewhat conspicuously, left out 2D and 3D heatmaps and contour plots, and similar, from the conversation. This isn't without reason as they will get their own whole part, which is coming next. Stay tuned!

Resources

• The unspoken rules of visualisation
• violin plots should not exist
• DataViz Project
• Questions in DataViz
• The Visual Display of Quantitative Information by Edward R Tufte
  • If you want to buy a textbook on DataViz, buy this one. It is the BEST.

The Series

This is part of a series of four articles on the importance of quality data visualization. Find links to the rest of the series here (as they are published):

• Data Visualisation is Very Important and you should care about it: Part 1
• Data Visualisation is Very Important and you should care about it: Part 2
• Data Visualisation is Very Important and you should care about it: Part 4