DRAFT This draft is under development — DO NOT CITE OR SHARE.

3  Basic Definitions

3.1 Reproducibility (Computational)

Reproducibility means that someone else can take the same data, the same code, and the same computational environment and obtain the same results. In this handbook, reproducibility is a practical goal: you should be able to rerun a method, verify its output, and reuse it for your own data with minimal friction.

3.2 Replicability (Independent Confirmation)

Replicability means that independent researchers can confirm a finding using new data (and often different code). Replicability is a scientific standard, while reproducibility is a computational standard. Both matter, but they are distinct.

3.3 Literate Programming

Literate programming is a style of writing where code and explanation are interwoven. The document is produced by executing the code, so the narrative and the results remain consistent by construction.

3.4 Reproducible Computing

Reproducible computing is the practical application of reproducibility in statistical work: data, code, and results are connected in a way that can be rerun, verified, and adapted by the reader.

3.5 Research Compendia (Modern View)

A research compendium is an organized package of text, code, data, and environment details that enables reproduction and reuse. In modern practice, this includes:

  • the analysis script or notebook,
  • the data or data source,
  • the computational environment (e.g., package versions),
  • the resulting outputs (figures, tables, reports).

This handbook itself is a web-native compendium: it combines explanatory text, code-backed results, and interactive computational tools in a single, navigable system.

3.6 How This Handbook Implements Reproducibility

This handbook is built around a reproducible workflow:

  • Quarto generates executable documents in which results are tied to code.
  • Shiny apps expose methods as interactive computations that can be reproduced and adapted by the reader.
  • Links and embedded inputs make it easy to jump from the explanation to a working computational implementation.

The intent is not only to explain methods but to make them directly usable and recomputable.

3.7 Minimal Reproducibility Checklist

For any method or analysis, ask:

  1. Can I access the data (or an equivalent example)?
  2. Can I see or run the code?
  3. Is the computational environment specified or reproducible (e.g., R/package versions, renv lockfile, or container)?
  4. Are the results traceable back to the exact inputs and code used (e.g., a Shiny URL with embedded inputs or an archived computation)?

If the answer is “yes” to all four, the analysis is reproducible.

3.8 Historical Note: The R Framework and Compendium Platform

The original vision of this handbook came from the R Framework and the Compendium Platform, which aimed to make statistical computation reproducible on the web. The R Framework provided modular statistical services, while the Compendium Platform stored computations as archived, citable resources. This early infrastructure influenced the handbook’s emphasis on transparency, reuse, and traceable computation.

While today’s implementation uses more modern tooling (Quarto, Shiny, and web-based deployment), the goal remains the same: make statistical analysis open, reproducible, and immediately usable by the reader.