Skip to content

Hadley Wickham

Abstract

Hadley Wickham (born 1979) is a New Zealand statistician who never sat on the R Core Team and changed the language more than most of the people who did. Starting as a doctoral student at Iowa State, he wrote packages that replaced the parts of R people found ugliest: ggplot2 (2007) for graphics, built on a formal grammar; reshape and plyr for reshaping tables; dplyr (2014) for the verbs of data manipulation. In 2014 he gave the underlying idea a name, tidy data, and in 2016 collected the packages as the tidyverse. From 2013 he did this as chief scientist of RStudio, now Posit, a company that gave the software away and sold the tools around it. Base R still exists underneath; most people who learned R after 2016 learned his dialect first.

Hadley Wickham
Hadley Wickham, 2016. Image: Hadley Wickham, CC BY-SA 4.0, via Wikimedia Commons.

Hamilton

Hadley Alexander Wickham was born in Hamilton, New Zealand, on October 14, 1979. At the University of Auckland, where R had been written a few years earlier, he took a bachelor’s degree in human biology and a master’s in statistics between 1999 and 2004, then went to Iowa State University for a doctorate under Dianne Cook and Heike Hofmann, two of the field’s leading researchers on statistical graphics. His thesis, finished in 2008, was called Practical Tools for Exploring Data and Models, and the tools were R packages. He had already won the American Statistical Association’s John Chambers Award for statistical software in 2006, as a student, for the first two of them, ggplot and reshape.

A Grammar

The package was ggplot2. Leland Wilkinson’s 1999 book The Grammar of Graphics had argued that a statistical graphic is not a chart type chosen from a menu but a composition: data, mapped by aesthetics to geometric objects, arranged by scales and coordinate systems, split into facets. Wickham implemented that grammar in R, first as ggplot, released on CRAN in 2006, and then as the rewrite ggplot2 in 2007. A user did not ask for “a scatterplot” but declared that x maps to one column, y to another, colour to a third, and asked for points; a different geometry or a facet was one more clause. The result was graphics that were consistent, layered, and slower than base R’s, and that people were willing to wait for. The book followed in 2009.

The second set of packages was about the shape of tables. reshape (2005) and plyr (2008) handled the operations analysts spent most of their time on, splitting data by groups, applying a function, and reassembling it, and in 2014 he stated the principle underneath them in a Journal of Statistical Software paper, “Tidy Data”: each variable is a column, each observation is a row, each type of observational unit is a table. It was Codd’s third normal form restated for statisticians, and it gave a name to the format the tools expected. dplyr, released the same year with Romain FranΓ§ois, reduced data manipulation to a few verbs, filter, select, mutate, summarise, arrange, chained with a pipe, and ran them fast enough on large tables to matter.

The Tidyverse

He had been an assistant professor at Rice University in Houston from 2008 to 2012. In 2013 he became chief scientist at RStudio, the company whose integrated development environment most R users worked in, and the arrangement defined the rest of his career: the packages were free and open, the company sold servers and support, and Wickham’s job was to write the software and teach it. Users had started calling the growing collection the “Hadleyverse”; he disliked the name, and at the useR! conference in June 2016 he announced the tidyverse, a meta-package released that September that installed the core set, ggplot2, dplyr, tidyr, readr, purrr, tibble, stringr and forcats, under one philosophy and one style. By 2018 its packages were among the ten most downloaded on CRAN. RStudio renamed itself Posit in 2022 to signal that it served Python users too.

The books did as much as the code. Advanced R (2014) explained the language’s semantics, including the parts inherited from Scheme that few users understood; R Packages (2015) turned package authorship from folklore into a procedure; R for Data Science (2017, with Garrett Grolemund) became the standard introductory text and is free online. All were written in R and published on the web before they were printed.

The tidyverse has critics. It is a second dialect that a base-R programmer must learn, its non-standard evaluation makes some ordinary programming tasks harder, and it puts the shape of a large part of the R ecosystem in the hands of one company. Wickham’s answer has been to publish the design principles, keep the licences open, and keep shipping. He was made a Fellow of the American Statistical Association in 2015 and received the COPSS Presidents’ Award, the field’s prize for a statistician under 41, in 2019, cited for statistical computing, visualisation and data analysis. He holds adjunct appointments at Auckland, Stanford and Rice.

πŸ“š Sources