Reproducibility

Updated Mar 04, 2023 ·

Overview

For reliable results, it's important to ensure your analysis is both reproducible and replicable.

• Reproducibility

  • Running the same analysis should produce identical results.
  • Ensures consistency in findings.

• Replicability

  • Use different tools but follow the same process.
  • Shows methods can be adapted and reused.

Benefits

These qualities help prevent wasted effort and allow others to build on your work.

• Peer Review

  • Enables others to verify and improve your work.
  • Promotes collaboration and validation.

• Scientific Progress

  • Builds on existing work and drives innovation.
  • Encourages ongoing learning.

Best Practices

1. Maintain clarity and track your changes.

   • Comment code and list used packages.
   • Helps others understand your process.
   • Keeps a record of modifications for easy revision.

2. Avoid manual data manipulation.

   • Use tools like Git to track changes.
   • Keep different dataset versions.
   • Tracks changes and makes transformations clearer.

3. Control randomness for consistency.

   • Set a random seed to ensure consistent results.
   • Control confounding variables.
   • Ensures model changes are deliberate.

4. Make your work easy to understand.

   • Present findings clearly for stakeholders.
   • Helps others replicate and understand your process.

5. Always credit the original sources.

   • Use formats like APA to cite sources.
   • Ensures proper referencing and easy access.

Manage References

Use tools to manage citations.

• Simplifies citation management and style switching.
• Tools like EndNote, Mendeley, and RefWorks.