Reproducibility

Reviewing a result requires knowing how data, processing settings, and figures fit together. We distinguish what can be rerun from what can only be examined through the available documentation.

DATA & METHODS5 SECTIONS

Understand the Available Level of Reproduction

Exact regeneration, a repeat analysis under comparable conditions, and review of a documented procedure require different information. Where data or code cannot be released, consider how much of the result can be checked with the information available.

Trace Data, Settings, and Figures

The following information helps with review:

  • input-data sources, versions or access dates, and target intervals
  • preprocessing and analysis settings, run steps, and major dependencies
  • identifiers and versions for input, intermediate, output, and figure files
  • randomness or manual decisions that need to be matched

Check What Changed

Data updates, code corrections, and changes to figures or wording can affect how a result appears or is explained. Dates, versions, and reasons for changes help readers compare it with an earlier result. Separating observations from interpretation also supports review.

Discuss Reanalysis or Collaborative Review

For a reanalysis inquiry, please include the dataset, processing settings, figure version, and the question you want to examine. What can be shared depends on usage terms, collaboration agreements, and confidentiality.

Explore Software EnvironmentDiscuss Reanalysis