Reproducible Research

• Reproducible research is a cornerstone of modern data science.
• The biggest enemy of reproducible research is complexity.
• Taming complexity requires a foundation (i.e., a framework) that provides a structure for the development of the analytic capacities providing for an easily maintained integration of the analytic tools, documentation, and content associated with the analytic tools.
• Without a solid foundation, analytic tools often become “spaghetti code” that is barely reusable, not reproducible, and not well documented.
• The goal of packageSkeleton is to provide a foundation for data science research that is reproducible, reusable, and well documented.

Introduction to packageSkeleton

• packageSkeleton is a streamlined GitHub template designed to simplify R package development.
• It also integrates with a Quarto-powered website hosted on GitHub, providing users a structured way to develop, document, and distribute their work.

Why packageSkeleton?

• Encourages best practices in R package development
• Provides a structure that integrates R code and documentation
• Simplifies setting up a website for sharing your work
• Ideal for both beginners and experienced developers

Key Features

• R Package Template: A basic structure that you can easily customize for your package needs.
• GitHub Integration: Fork or clone the repository and start developing right away.
• Quarto Documentation: Create rich, dynamic content in multiple formats (e.g., HTML, PDF, DOCX, PPTX, etc.) that documents your package and provides a public-facing website.

Customizing packageSkeleton

1. Rename the Package: Start by modifying the DESCRIPTION file and renaming the package.
2. Add Your Own Functions: Inside the R/ directory, include your custom R functions.
3. Generate Documentation: Use roxygen2 to create function documentation and Quarto to build your website.
4. Deploy the Website: Publish the Quarto website using GitHub Pages.

Requirements

• R: The latest version of R (available from CRAN)
• Quarto: The Quarto CLI tool for building documents and websites (Install Quarto).
• Git: Version control to manage your code and push to GitHub.
• GitHub: Host your R package and website.

GitHub Workflow

1. Fork the Repository: Start by forking the packageSkeleton repo from GitHub.
2. Clone Locally: Clone the repo to your local machine using Git.
3. Push Changes: After making your edits, push them to your GitHub repository.
4. Deploy Website: Use GitHub Pages to host the Quarto-based documentation.

Customizing the Website

• Quarto Configuration: Customize the _quarto.yml file to change the website settings such as title, author, and repository URL.
• Add Pages: Create .qmd files to add custom documentation, tutorials, and more.
• Themes & Layouts: Modify the website’s theme using CSS or choose from Quarto’s built-in themes.

Building Your Package

• Add Functions: Use the R/ folder to store your custom R scripts and functions.
• Generate Documentation: Write roxygen-style documentation for each function and build it using devtools::document().
• Check Your Package: Ensure your package passes all R CMD checks by running devtools::check().

Publishing the Website

1. Build the Website: Run quarto render to build the static HTML content.
2. Publish on GitHub Pages: Push your changes to the repository and GitHub Pages will automatically deploy the site.
3. Preview Locally: Use quarto preview to see live updates of the website as you edit.

Conclusion

• packageSkeleton is a powerful starting point for R package development.
• Use it to streamline your workflow, create well-documented R code, and publish it with a professional Quarto-powered website.
• Leverage GitHub for collaboration, version control, and public distribution of your work.

Future Directions

• NextJS Integration: Integrate packageSkeleton into NextJS framework to facilitate the development of a more robust and scalable data science development framework.
• Web API Development: Expose analytics results as web APIs to be used by users as well as other applications.
• MCP Integration: Utilize Model Context Protocol (MCP) to expose package tools and resources for integration with popular AI services.

Resources

• Quarto Documentation
• R Package Development
• GitHub Pages

For the Presentation