Commonality inferences

Code and data for reproducing the analyses in the paper.

Requirements

• Python 3.10+ with scientific packages (numpy, pandas, scipy, matplotlib, seaborn, jax)
• R 4.0+ with packages: lme4, lmerTest, dplyr, tidyr, reticulate
• Quarto 1.3+: https://quarto.org/

Setup

Option 1: Using uv (recommended)

uv is a fast Python package manager that handles virtual environments automatically.

# Install Python dependencies (creates .venv automatically)
uv sync

# Install R packages
Rscript -e "install.packages(c('lme4', 'lmerTest', 'dplyr', 'tidyr', 'purrr', 'reticulate'))"

Option 2: Using pip

# Create and activate a virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install Python dependencies
pip install -r requirements.txt

# Install R packages
Rscript -e "install.packages(c('lme4', 'lmerTest', 'dplyr', 'tidyr', 'purrr', 'reticulate'))"

Reproducing Analyses

The Quarto notebooks use R's reticulate package to run Python code. You must tell R which Python to use by setting the RETICULATE_PYTHON environment variable.

Quick start

# Set Python path and render all notebooks
RETICULATE_PYTHON=$(pwd)/.venv/bin/python quarto render analysis/model_analyses.qmd
RETICULATE_PYTHON=$(pwd)/.venv/bin/python quarto render analysis/behavioral_analyses.qmd
RETICULATE_PYTHON=$(pwd)/.venv/bin/python quarto render analysis/supplement.qmd

Alternative: Export once per session

# Export the variable for your shell session
export RETICULATE_PYTHON=$(pwd)/.venv/bin/python

# Then render notebooks without prefix
quarto render analysis/model_analyses.qmd
quarto render analysis/behavioral_analyses.qmd
quarto render analysis/supplement.qmd

Individual notebooks

Notebook	Description	Main outputs
model_analyses.qmd	Bayesian factor model	Figure 4, Figure 5
behavioral_analyses.qmd	Mixed-effects models	Figure 2, Figure 3
supplement.qmd	Supplementary analyses	SI figures and tables

Troubleshooting

"Python specified in RETICULATE_PYTHON does not exist"

• Make sure you ran uv sync or pip install -r requirements.txt first
• Check the path: ls -la .venv/bin/python
• Use absolute path: export RETICULATE_PYTHON=/full/path/to/shared-reality/.venv/bin/python

R can't find packages

• Install missing R packages: Rscript -e "install.packages('package_name')"

JAX errors on Apple Silicon

• JAX should work out of the box on M1/M2 Macs with the dependencies specified

Structure

├── analysis/           # Quarto notebooks
│   ├── model_analyses.qmd      # Bayesian factor model (Figure 4, 5)
│   ├── behavioral_analyses.qmd # Mixed-effects models (Figures 2, 3)
│   └── supplement.qmd          # Supplementary analyses
├── data/               # Experimental data
│   ├── responses.csv   # Main behavioral data
│   ├── questions.csv   # Survey questions
│   └── llm_results/    # LLM prediction outputs
├── models/             # Computational model code
│   ├── model.py        # Bayesian factor model
│   └── llm/            # LLM prediction pipeline
├── outputs/figures/    # Generated figures
├── pyproject.toml      # Python dependencies (for uv)
└── requirements.txt    # Python dependencies (for pip)