Dos and Don’ts of Data Visualisation
Whether on purpose or not, a statistician can mislead an audience with a chart. This article explains some chart design principles and common mistakes novice data analysts make. Think about the statistical charts you have seen on billboards, in the news, and in research studies. Using these principles as a guide, would you classify any of those charts as misleading? Be sure to take note of the suggestions for successful dashboards.
Make charts easy to read
Don't use more than (about) six colours
Colour categorisation is not random, and the centres of basic colour terms are very similar in all languages. Using colour categories that are relatively universal makes it easier to see differences between colours. The figure below shows the order of appearance of colour names in languages around the world. The order is fixed, with the exception that sometimes yellow is present before green and sometimes the reverse is the case.
Use different colours to represent different categories (e.g., private/public, types of pollutant), not different values in a range (e.g., age, temperature). See qualitative colour palette below.
If you want colour to show a numerical value, use a range that goes from light to dark in one of the universal colour categories. See sequential colour palette below.
If you need to represent diverging numeric values (from hot to cold, from good to bad, etc.), use two colours as shown in the diverging colour palette example.
Do not use rainbows for range values.
