Making my first data package

packages
methods
Author

Sharon Howard

Published

24 June 2025

R Packages

Packages, including data packages, are a vital part of the R ecosystem and yet I’ve been using R for the best part of a decade without ever making one of my own. Until now.

I’ve recently been reading about using R data packages for reproducible research, and I decided it was time to rectify my omission by making a small practice package of cleaned-up datasets for reuse on this blog.

There are plenty of resources (see list at the end of the post), but some are a lot more detailed than you need to get going. I found this step by step introduction particularly helpful. I also found it useful to look at the structure of existing data packages such as historydata and HistData

Step by Step

This is largely based on the guide linked above but it’s not a full set of instructions! There are several steps to the process; they aren’t difficult but they need to be done in the right order. So this is really just a checklist to guide me through in future.

  1. Think of a good name for the package (there are strict rules about naming)
  2. Check you have necessary packages installed (devtools, usethis, roxygen2)
  3. Create new project as an R package in RStudio (I recommend ticking the git repo box) or with usethis::create_package(). It will contain various mandatory files and folders (including R/, data/ and man/)
  4. Put datasets in the data/ folder; these must be .rda (R data) files, one per dataset
  • I put raw data files in an additional data-raw folder with some .R scripts for processing. Extras like these shouldn’t go in the data/ or R/ folders. To make the actual package data, need an extra line at the end of the processing script: usethis::use_data(dataset_name, overwrite = TRUE). That will create the required .rda file in the data/ folder; to update it you’d just run the script again.
  • Or you can just make .rda files in another project and copy them into data/
  1. Make an R script for each dataset in the R/ folder, written in a special format where each line begins with # ' (I copy and pasted from existing data packages to get started).
  2. Create package metadata: DESCRIPTION and license.
  • For a data package you might want a Creative Commons license, eg usethis::use_ccby_license()
  1. Load documentation with devtools::document() (this turns the step 5 R scripts into .Rd files in the man/ folder)
  2. Do checks in RStudio Build tab, or with devtools::check()
  3. Install the package locally, in RStudio Build tab under Install, or devtools::install()
  • repeat this step if you make any changes to the package
  1. Publish to Github, if you want to make it shareable.

The datasets

The package (though it’s only really intended for my use and lacks proper documentation) is on Github. It may expand and get properly documented in the future.

I started with just a couple of datasets that I’ve previously explored here or elsewhere, coroners (London Lives coroners’ inquests) and cheshire_petitions (TPOP Cheshire petitions).

Code 
library(mindseyedata)
library(dplyr)
library(ggplot2)
library(waffle)
library(ggthemes)
Code 
coroners |>
  count(verdict, gender) |>
  filter(verdict !="undetermined") |>
  ggplot(aes(fill=verdict, values=n)) +
  geom_waffle(colour="white", n_rows = 10, make_proportional = TRUE) +
  coord_equal() +
  scale_fill_ptol() +
  theme_minimal() +
  theme_enhance_waffle() +
  facet_wrap(~gender) +
  ggtitle("Westminster coroners' inquests: verdicts by gender")

Code 
cheshire_petitions|> 
  add_count(topic, name="n_topic") |>
  filter(n_topic>25) |>
  count(petition_type, topic) |>
  ggplot(aes(fill=petition_type, values=n)) +
  geom_waffle(colour="white", n_rows = 10, make_proportional = TRUE) +
  coord_equal() +
  scale_fill_pander() +
  theme_minimal() +
  theme_enhance_waffle() +
  facet_wrap(~topic) +
  ggtitle("Cheshire petitions types and topics")

Resources