useR! 2024

With the rise of web technologies, interactive data visualizations have become a staple of data presentation. Yet, despite their growing popularity, researchers still point to the lack of a formalized pipeline for turning raw data into summary statistics. The cause of this lack may be a subtle yet profound issue: while we often treat statistics and graphical objects as independent, they are in fact deeply connected. Consider a typical stacked barplot. Many researchers have noted that stacking some summaries will produce a valid overall statistic (e.g. count, sum), whereas stacking others will not (e.g. mean). But what are the mathematical properties that make this possible? Are there other operators that can be stacked? The goal of this talk is to delve into the relationship between graphical objects, statistics, and interaction. Specifically, by discussing a handful of concepts from category theory, I hope to give you a new appreciation of the rich structure that lies beyond the figures we look at every day. Finally, this talk will also briefly introduce a new R package for interactive data exploration – plotscaper – which is an attempt to implement some of these ideas.