visualisation tips

Make set intersection great again with UpSetPlot

Venn Diagrams are dead, long live Venn Diagrams

Covid 19 symptoms, from http://upset.plus/covid19symptoms/

Venn diagrams are fantastic to communicate with. They convey very simply the notion of an entity belonging to multiple categories. Let’s say we are analysing purchase data and we’re discussing the fact that customers of item A are more likely to return than customer of item B. We may draw a diagram like the one below, with numbers associated to each area:

This makes communication easy, but there are many problems.

• With items of fixed shape, the size of intersections are interconnected. One cannot move or resize one shape without changing all its intersections.
• Areas are often not proportional to size.
• It can manage a maximum of 5 sets before being unreadable. (cf wiki)

Example from ggVennDiagrams R library

The diagram above shows these issues. Note the color issues, the zones where size is inverse…