Introduction
This website is the result of a project made during a teaching unit done by the FRB-CESAB in Montpellier (December 2021). All informations given in this website are based on the toolbox done for the teaching.
This website is about coding in a reproducible workflow using R and is especially made for ecological studies but most concepts can be used for any study.
What is Reproducible Research?
Doing a reproducible research is providing a “clean, repeatable script-based workflow” Bristish Ecological Society 2018 linking raw data to final analysis and outupts. It is important to follow a reproducible workflow so that other people can re-run your code or that a future version of you can easily redo the analysis!
It relies on several important points that are explained in the different tutorials (cf Tutorials and examples down below):
Project organisation: you have to correctly organise your working directory by creating a Research Compendium which have defined directories (cf Let’s Create a Research Compendium tutorial)
Data management: you have to keep raw data and save intermediate data. Never overwrite raw data!
If raw data you used is download, it is better if you automate their obtention (usingdownload.file()function for instance).Softwares - Scripts: you have to respect the agreements of the community you work in, for instance style rules on objects names or scripts aesthetic (example: tidyverse rules). One basic rule when writing R scripts is for instance to always use a relative path (where the file is according to the working directory)
Softwares - Environment: you have to indicate the version of used packages (cf the renv package using the Create and isolted and portable environment using renv tutorial) because some functions/parameters can change between packages versions. You can also track the version of R you are working on or the Operating System version (using for instance the Docker package cf Introduction of Docker for R Users by Colin Fay)
Softwares - Workflow: scripts succession can be automated using a
make.Rfile (cf Let’s Create a Research Compendium tutorial) or the target package: (cf Create an automatised workflow tutorial)Tracking Changes: using a version control software as Git (cf Use Git and Github tutorial).
Collaboration: you have to use a place to exchange to share your code, where you can collaborate with others and where other people can suggest modifications or improvement of your code throught Github or GitLab (cf Use Git and Github tutorial). To be able to collaborate with others you also have to document your project, notably by using a
READMEfile (cf Let’s Create a Research Compendium tutorial)Manuscripts: you can automate the writing process by using literate programming combining writing, code and figures using Rmarkdown
Tutorials and examples
In this website, we propose a set of tutorials to learn how to do reproducible research!
Each tutorial contains R chunks so you can run the analysis using a Pokemon dataset from the Kaggle website (link) which gathers informations like Basic Statistics, Type, Height, Weight etc for 802 Pokemons. We will create a reproducible project to answer two questions:
- How does the attack ability evolve with height and weight?
- Which pokemon type has the highest proportion of legendary pokemons?
We used R and RStudio (so screenshots are made under Rstudio). So that the tutorials correctly work, please be sure that you have a R version >= 4.0.0, that Git, Github are configurated to communicate with R and that your computer have everything needed to read Rmarkdown file. Here are instructions to do so (in french)
If you are new to the subject, you can follow the tutorials in this order:
1 - Let’s Create a Research Compendium!
2 - Use Git and Github
3 - Create and define your first R function and analysis script
4 - Create an automatised workflow: using make.R
Now let’s do reproducible research!
Pikachu Artwork from Pokepedia