Reproducible Research Practices

Day 1 | WISE Workshop

Sean Sylvia

Why Reproducibility Matters

Science requires verification
Your future self will thank you
Collaboration becomes possible
Errors are easier to find

The Reproducibility Crisis

“More than 70% of researchers have tried and failed to reproduce another scientist’s experiments.”

— Nature Survey, 2016

What Makes Research Reproducible?

Code - All analysis steps documented
Data - Raw data preserved and accessible
Environment - Software versions recorded
Documentation - Clear explanations

Project Structure

my-project/
├── data/
│   ├── raw/           # Never modify
│   └── processed/     # Derived data
├── notebooks/         # Analysis notebooks
├── src/               # Reusable code
├── outputs/           # Results, figures
├── README.md          # Project overview
└── requirements.txt   # Dependencies

Version Control with Git

Track changes over time
Collaborate without conflicts
Revert mistakes easily
Document your reasoning

Google Colab for Reproducibility

Advantages:

Consistent environment
Easy sharing
No setup required
Free compute resources

Best Practices

Never modify raw data
Document everything
Use relative paths
Record package versions
Write code for humans

Hands-on: Setting Up Your Project

Let’s create a reproducible project structure together.