1 Overview

This is a minimal introduction to reproducible research, including data management and handling data in R, compiled for the Biological Sciences BSc(Honours) class at the University of Cape Town.

This is a living document, and I will be updating it as we go along. It is written in R Markdown and compiled using bookdown. The source code for this document is available on GitHub and is licensed under a Creative Commons Attribution 4.0 International License. You are free to use and adapt the material as you see fit, but please provide attribution to the original author.

1.1 General

This really is a minimalist introduction. We only have a week! I’ll focus on providing a broad overview of the general framework and motivation for reproducible research (including good data management), teaching a few practical skills along the way.

Mostly this is not fun and exciting, but it is important stuff for any biologist to know. I’ll try my best to make it interesting! Hopefully by the end of the module you’ll see the value in it all - both for you as an individual and for science and society in general.

“Let us emphasize again this obvious conclusion: a scholar’s positive contribution is measured by the sum of the original data that [they] contribute… Hypotheses come and go but data remain. Theories desert us, while data defend us. They are our true resources, our real estate, and our best pedigree. In the eternal shifting of things, only they will save us from the ravages of time and from the forgetfulness or injustice of [people]. To risk everything on the success of one idea is to forget that every fifteen or twenty years theories are replaced or revised. So many apparently conclusive theories in physics, chemistry, geology, and biology have collapsed in the last few decades!” - Santiago Ramón y Cajal, 1906 Nobel Laureate, from Advice for a Young Investigator 1898 (Ramón y Cajal 1999)

i.e. ideas are cheap, data are priceless

1.1.1 Learning Objectives

The core outcomes/concepts I hope you’ll come away with:

  • Familiarity with the concepts and understand the need for Open, Reproducible Science
  • Familiarity with The Data Life Cycle
  • Some data management and handling skills

1.2 Lectures/Discussions/Tutorials

These will be held live in person in BIO LT1 from 9:30 to 11:30AM from the 17th to the 21st February unless otherwise announced on Amathuba.

I’ll be adding to (and mostly teaching from) these online course notes as we go along.

The schedule of lectures (and readings) is as follows:

There will also be an online UCT Data Management Plan tutorial with Sanjin Muftic (UCT Libraries) on Wednesday, 19th Feb at 1PM.

1.3 Deliverables (Due Thursday the 27th February)

  1. A draft Data Management Plan (DMP) for your Honours Project (or mock project) using UCT’s Online DMP Tool. Please use the “UCT Student Generic - Full DMP” template. This is a draft, so don’t worry too much about getting it perfect, but do try to think about what data you’ll be collecting, how you’ll store it, how you’ll back it up, and how you’ll share it. You’ll need to submit a link to the DMP with your Honour’s project, so it’s worth getting a head start on this now.

  2. A fully reproducible GitHub repository containing suitably named sub-folders, data files (if small) and the R scripts (that you’ll develop based on Wednesday’s tutorial), all in line with best practice as per the content of this module. It will also need have a Binder badge in the README that works and allows me to execute your code online (instructions below).

  • Again, best if these are data related to your Honour’s project (speak to your supervisor), but it can be a different dataset or a mock dataset if you prefer (e.g. you could simulate the kind of data you expect to be working with for your project).
  • More about the R script during the Tidy data tutorial, but it must be easily executable by Binder, and output your data in tidy format and a summary figure of some kind.
  • Since it will be public, it is best to censor or obfuscate your data if they are not open access. If your data are large (>5MB), then it’s best to create and only upload a smaller subset of the data. Note that your R script must still work with the reduced dataset, since part of your mark will be based on whether your script runs and the output reproducible.
  • To use Binder with R, you first have to add a runtime.txt file to your repository that specifies the version of R you want to use and an install.R file that specifies the R packages you want to install. You can find examples of these files at https://github.com/jslingsby/bogusrepo. Note that you’ll need to add any specific libraries you use in your R script to the install.R file.
  • The Binder badge can be created by going to Binder, pasting the URL of your GitHub repository into the “GitHub repository name or URL” box, and clicking “launch”. If it works, you’ll be taken to a page where you can interact with your repository online. Copy the URL of this page (under “Badges for your README” on the Binder landing page) and paste it into the README of your repository. This will create a badge that will take anyone who clicks on it to the Binder page where they can interact with your repository online. NOTE: Binder sometimes fails. I think this is because it’s a free service and gets overloaded. If it fails, try again later. If it fails repeatedly, try again later. If no luck before the submission deadline, let me know and submit the link to your repository without the Binder badge. It also sometimes freezes and looks like it’s running for ages (>15 minutes). In those cases it has usually worked and just not told you, so try the link.
  • Please test your repository by clicking on the Binder badge in your README. If it works, you should be taken to a page where you can interact with your repository online. Click on the RStudio icon to open an RStudio session in your browser. You can then open and run your R script (or RMarkdown or Quarto file) from here. If it doesn’t work, your repo probably isn’t reproducible. Check the warning and troubleshoot…
  1. Lastly, I’d like you to spend some time reading up about reproducible research and data management (preferably relevant to our field) and to share any interesting resources you find - a link to a paper, news article, blog, tutorial, etc - with a 50 to 200 word summary of what it is and why its relevant and interesting. These will be collected and shared with the class either via Google Forms or GitHub (just as soon as I’ve set it up…).

1.4 Software installation and setup (Due Monday 17th Feb)

For the data wrangling exercise and the second deliverable, we’ll be using the R statistical programming language and the Git version control system. We’ll also be using an integrated development environment (IDE) for each: RStudio and GitHub, respectively.

If you already have these installed and set up, please make sure you have the latest versions, and check that your installations are working! Please also make sure you have installed (and/or updated) the Tidyverse set of R packages. It can be installed using the code install.packages("tidyverse") and updated using update.packages("tidyverse").

The installation and setup can be a bit long-winded, but once done you should be good to go until you change or reformat your computer. The steps below are my summary and (hopefully) more intuitive adaptation of the instructions provided for setting up GitHub and version control with R. If my steps don’t work its probably best to read up there, or at Happy Git with R.


First we’ll start with the necessary software.

  1. Download and install the latest version of R
  2. Download and install the latest free version of RStudio Desktop
  3. Download and install the latest version of Git - accept all the defaults
  4. Create a GitHub account
  5. Run through the 10 minute GitHub tutorial that is offered when you activate your GitHub Account (It’ll really help you get the idea behind what Git does!)

Now you have RStudio, R and Git installed, and you have a working GitHub account that lets you do stuff online, but what remains is to get GitHub working locally and configuring RStudio to use GitHub. If any of the following doesn’t work, have a look at the installation section of Happy Git with R to troubleshoot:

  1. Install GitHub CLI (Command Line Interface).
  2. Open RStudio.
  • Select the Terminal tab (top left, next to Console)
  • Enter gh auth login, then follow the prompts:
    • Select GitHub.com
    • When prompted for your preferred protocol for Git operations, select HTTPS
    • When asked if you would like to authenticate to Git with your GitHub credentials, enter Y
    • When asked how you would like to authenticate select Login with web browser
    • Copy the 8-digit code and hit Enter
    • Github.com will open in your internet browser - paste the code and hit enter
    • If any of these steps don’t work, just start again with gh auth login in Terminal

Now that GitHub is starting to enforce 2-factor authentication, you may also need to do the following (especially if you get login issues at the end of step 10):

  • We need to add a “personal access token” (PAT) which is generated by GitHub and added to R. We’ll manage this using library(usethis) as it simplifies things. See this article for a full explanation.
    • First, install the package by running install.packages("usethis") in the R console
    • Second, create a Github authentication token by running usethis::create_github_token() in the R console
      • Follow the prompts. Give the token a sensible name like “personal-laptop-rstudio-git” so you can work out what it is if you need to later
      • Copy the code displayed - You will not be able to see it again!
    • Third, run gitcreds::gitcreds_set() in the R console
      • When prompted, paste your token into R
  1. In RStudio
  • Go to Global Options (from the Tools menu)
  • Click Git/SVN
  • Make sure Enable version control interface for RStudio projects is on
  • If necessary, enter the path for your Git or SVN executable where provided (this shouldn’t be needed, but may)
  • Click Apply
  • Restart RStudio

Lastly, you need to install the Tidyverse set of R packages. This can be done using the code install.packages("tidyverse").

References

Baker, Monya. 2016. “1,500 scientists lift the lid on reproducibility.” Nature 533 (7604): 452–54. https://doi.org/10.1038/533452a.
Markowetz, Florian. 2015. “Five selfish reasons to work reproducibly.” Genome Biology 16 (December): 274. https://doi.org/10.1186/s13059-015-0850-7.
Michener, William K, and Matthew B Jones. 2012. “Ecoinformatics: supporting ecology as a data-intensive science.” Trends in Ecology & Evolution 27 (2): 85–93. https://doi.org/10.1016/j.tree.2011.11.016.
Peng, Roger D. 2011. “Reproducible research in computational science.” Science 334 (6060): 1226–27. https://doi.org/10.1126/science.1213847.
Ramón y Cajal, Santiago. 1999. Advice for a young investigator. The MIT Press. https://doi.org/10.7551/mitpress/1133.001.0001.
Wickham, Hadley. 2014. “Tidy Data.” Journal of Statistical Software, Articles 59 (10): 1–23. https://doi.org/10.18637/jss.v059.i10.
Wickham, Hadley, Mara Averick, Jennifer Bryan, Winston Chang, Lucy McGowan, Romain François, Garrett Grolemund, et al. 2019. “Welcome to the tidyverse.” Journal of Open Source Software 4 (43): 1686. https://doi.org/10.21105/joss.01686.