Introduction to the tidyverse

Authors: Clark S. Rushing
Contributors: Michael T. Hallworth

What is the tidyverse?

If you have been using R over the last several years, you’ve probably noticed an ever-growing number of R enthusiasts extolling the virtues of the tidyverse, “an opinionated collection of R packages designed for data science.” The tidyverse is built around several core packages, including dplyr, tidyr, ggplot2, and purrr, plus many others that provide more specialized solutions for common data science problems. All packages in the tidyverse share a common philosophy about data structure, coding style, and workflow, which allows tidyverse functions to work together more-or-less seamlessly. Although not all R users agree with this philosophy or like using the tidyverse, the main advantages of the tidyverse include:

  • Consistent data input/output structure: The tidyverse is built around data frames. This means that, for the most part, tidyverse functions take a data frame as input and produce a new data frame as output. Perhaps more importantly, tidyverse functions are built around ‘tidy’ data frames - i.e., data frames where each row is a single observation and each column is a variable. Although not the most compact way to store data, tidy data has many advantages for manipulation, modeling, and visualization.
  • Intuitive function names: To the extent possible, tidyverse functions strive to have intuitive names so users can more easily understand what a function does. Examples include filter, arrange, summarise, and group_by.
  • Consistent style: Tidyverse functions use a common style guide. This means that naming conventions (e.g., functions names are verbs, all lower case with words separated by _), argument order (e.g., data is always the first argument), spacing and indentations, etc. are consistent across all tidyverse packages. As Hadley Wickham put it, this consistency “is like correct punctuation: you can manage without it, butitsuremakesthingseasiertoread”.
  • Piping functions: Rather than nesting functions (e.g. head(mean())), tidyverse functions can be sequentially linked using the pipe operator %>%. This greatly increases readability of code.