📊 A Probabilistic Grammar of Graphics

Turning a probability distribution into a visualization can be difficult and even error-prone, a lesson I learned from googling and overwriting default arugments to geom_density in ggplot2. The Probabilistic Grammar of Graphics promises to make our lives easier by using probability expressions like P(A|B) as data variables. The design of this GoG extension also corresponds nicely to uncertainty communication theories, and may help explore and formalize the space of uncertainty visualizations.

I designed the PGoG grammar and implemented a proof-of-concept in R. Currently mentoring two undergraduate students to refine the R library.

A probabilistic visualization gallery

Part of the design process was just to collect lots of visualizations. Here’s the collection I built, with some additions from Puhe Liang. The full collection is at Notion.