A guide to modern reproducible data science with R
summary
Shows how to turn any analysis into a minimal reproducible DESCRIPTION file compendium (just metadata, README, license, data/ and Rmd/) and when to add a short Dockerfile (repo2docker or containerit generate one). Demonstrates piggyback to attach files up to 2 GB to GitHub releases, GitHub file limit ~150 MB, CRAN package data limit 5 MB and ~37% of CRAN packages ship data. Binder builds cached images and launches an RStudio server live; Drake replaces complex Makefiles for R workflows. Use Zenodo for long‑term archiving.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:05Research compendium
- 4:42Description file
- 6:05Setting up a compendium
- 8:20Managing data
- 9:21piggyback
- 10:36arcdb
- 11:14isolating your environment
- 12:32binderhub
- 13:23launch binder
- 14:37binder hub
- 15:05binder hub setup
- 17:05make files
- 18:12Drake
- 18:50Interactive dependency graph
- 19:17Take home messages
- 19:55Conclusion
related talks
24mHow to Write Reproducible Manuscripts
46mElegant R: Tips from a Software Developer
25mReproducibility Around a Web Application
22m. Cultivating Your Own R Ecosystem as a Solo Contributor
19mA Docker Container Toolbox for the Data Scientist
17mWhy is computational reproducibility important?
46mJump-Start Your Data Science Learning Java & Jupyter Notebooks
48mCommon Data Science Mistakes