Learning from reproducing computational results: introducing three principles and the Reproduction Package

Area of Science:

• Computational Science
• Scientific Computing
• Research Reproducibility

Background:

• Reproducibility is crucial for scientific integrity.
• Barriers to computational reproducibility hinder scientific progress.
• Existing methods for sharing computational research are insufficient.

Purpose of the Study:

• Identify barriers to computational reproducibility.
• Propose principles and guidelines for reproducible research.
• Introduce a novel 'Reproduction Package' for sharing computational artifacts.

Main Methods:

• Attempted to reproduce computational results from seven published articles.
• Derived three core principles for reproducible research.
• Developed 12 specific guidelines for implementing these principles.

Main Results:

• Identified key barriers to computational reproducibility.
• Formulated three guiding principles: transparency, ease of (re-)executability, and deterministic code.
• Defined a novel 'Reproduction Package' formalism for sharing research artifacts.

Conclusions:

• Adherence to the three principles enhances computational reproducibility.
• The 'Reproduction Package' provides a structured method for sharing artifacts.
• Publicly available reproduction efforts serve as exemplars for the scientific community.