Introduction to Statistical Causal Inference

This repository contains the material for the introduction seminar to Statistical Causal Inference (SCI). The purpose of the seminar is to introduce software engineering researchers with a background in analysis of quantitative data to a causal framework for inferential statistics proposed by Judea Pearl¹ and Richard McElreath.²

Versions and contributors

Version	Date	Occasion	Contributors	Material
v1.0	2024-10-18	Seminar at UPC, Barcelona	Julian Frattini	2024-10-18-seminar-upc.pdf
v2.0	2025-04-28	Tutorial at the RE'25 conference	Julian Frattini, Hans-Martin Heyn, Robert Feldt, Richard Torkar	2025-04-28-tutorial-reconf.pdf
v2.1	2026-01-20	Guest lecture and seminar at TUM Heilbronn, Germany	Julian Frattini	2026-01-20-guestlecture-tum.pdf and 2026-01-20-seminar-tum.pdf
v2.2	2026-03-18	Seminar in Robert Feldt's research group	Julian Frattini	2026-03-18-seminar-feldt.pdf

Structure

The repository contains the following directories and files.

├── publicity : advertising material for the seminar
│   ├── banners: image files (from PowerPoint slides) for social media posts
│   ├── debriefing: summary of tutorial instances
│   └── screenshots: images from the slides for tutorial applications
├── slides : PowerPoint presentations for teaching the tutorial/seminars
│   ├── pdf: animation-free export of the presentations to PDF
│   └── presentations: complete slide decks for presentation
│       ├── 2024-10-18-seminar-upc.pptx: complete introduction to both SCI and BDA
│       ├── 2025-04-28-tutorial-reconf.pptx: conference tutorial introducing to SCI
│       ├── 2026-01-20-guestlecture-tum.pptx: guest lecture introducing to SCI
│       ├── 2026-01-20-seminar-tum.pptx: guest lecture focusing on causal modeling
│       └── 2026-03-18-seminar-feldt.pptx: seminar focusing on DAGs
└── src : source code to follow along the examples
│   ├── basics : description of fundamental concepts
│   │   ├── regression.Rmd : demonstration of the basic statistical analysis tool
│   │   └── simulations.Rmd : demonstration of ground truth simulations
│   ├── bda : implementations of BDA concepts and techniques
│   │   ├── brms : code snippets using the brms package
│   │   │   ├── bda-complete.Rmd : complete example of a simple Bayesian regression model
│   │   │   └── nonlinear.Rmd : estimation of a non-linear effect with different models
│   │   └── rethinking : code snippets using the rethinking package
│   │       ├── prior-predictive-checks.Rmd : demonstration of prior predictive checks 
│   │       └── model-notation.Rmd : demonstration of statistical model specification
│   ├── exercises : collections of exercises to test the acquired skills
│   │   └── exercise-d-separation.Rmd : collection of exercises in identifying adjustment sets
│   ├── sci : implementations of SCI concepts and techniques
│   │   ├── associations : explanation of the fundamental relationships between three variables
│   │   │   ├── collider.Rmd : demonstration of a common effect
│   │   │   ├── confounder.Rmd : demonstration of a common cause
│   │   │   └── mediator.Rmd : demonstration of a pipe
│   │   ├── dag.Rmd : demonstration of causal modeling with directed, acyclic graphs
│   │   └── model-comparison.Rmd : demonstration of model comparison to identify appropriate causal models
│   └── util : utility files and scripts with reused functions
│       └── extract-coefficients.R : script to extract all coefficient distributions from two models
└── sci-intro.Rproj : project file to open the project in RStudio

System Requirements

In order to run the R scripts and Rmd notebooks in the src folder, ensure that you have R (version > 4.0) and an appropriate IDE like RStudio installed on your machine. Then, ensure the following steps:

1. Install the C toolchain by following the instructions for Windows, Mac OS, or Linux respectively.
2. Restart RStudio and follow the instructions starting with the Installation of RStan
3. Install the latest version of stan by running the following commands

    install.packages("devtools")
    devtools::install_github("stan-dev/cmdstanr")
    cmdstanr::install_cmdstan()

4. Install all required packages via install.packages(c("tidyverse", "ggdag", "brms", "marginaleffects", "patchwork")).
5. Create a folder called fits within src/ such that brms has a location to place all Bayesian models.
6. Open the sci-intro.Rproj file with RStudio which will setup the environment correctly.

License

Copyright © 2024 Julian Frattini. This work is licensed under the Apache-2.0 License.

Footnotes

1. Pearl, J., & Mackenzie, D. (2018). The book of why: the new science of cause and effect. Basic books. ↩

2. McElreath, R. (2018). Statistical rethinking: A Bayesian course with examples in R and Stan. Chapman and Hall/CRC. ↩