Background

Multiple imputation is a powerful tool in presence of missing data.

However, especially in combination with propensity score analyses, multiple imputation can lead to challenges since analytic workflows can be become much more complex.

Objective

This repository showcases and evaluation different multiple imputation > propensity score > outcome analyses and associated implementation challenges.

Dependencies

This is a quarto book project and R package dependencies are listed in the _requirements.sh file, grouped by topic. Packages are installed with pak, including the development version of the encore.analytics package from GitHub (dev branch). To install all packages, run the following line of code in the Terminal:

sh _requirements.sh

Packages are installed into a project-local library (.rpkg/, set up by .Rprofile) and are not version-pinned, i.e., the latest available versions are installed.

Reproducibility

Follow these steps in RStudio to reproduce this study:

1. Clone this repository via git clone <url>
2. Install all necessary dependencies (see above)
3. Run all scripts via quarto render or (optionally) in RStudio Build > Render Book (make sure quarto is installed)

The book is rendered and deployed to GitHub Pages (gh-pages branch) by the Quarto Publish GitHub Action on every push to main.

The data used in this project is strictly simulated and no real patient-level data is used.