Reproducibility toolkit
What it will cover
- git for scientists: commits, branches, what to commit and what to ignore
- Environments: conda/mamba, uv, lock files, and recording versions
- Project layout: data/, notebooks/, src/, tests/, results/
- Seeds, logging and recording the exact inputs of every result
- Turning a notebook into a script and a script into a package
- Sharing: README, licence, citation, and a one-command rebuild
Planned learning outcomes
- Keep a project under version control and explain every commit.
- Reproduce a result from a clean clone with one command.
- Package a small analysis so that a colleague can run and extend it.
Until then, the reproducibility checklist in episode 2.6 and the Setup episode cover the essentials.