This repository contains code for the decision making course exam (Cognitive Science Aarhus University 2026). I constructed a bayesian model in PyMC and fit that to the data obtained in the study by Albert et al. (2024).

Dataset: Albert, L. P. (2024). Numerosity estimation of virtual humans as a digital-robotic marker for hallucinations in Parkinson's disease [Data set]. In Nature Communications (Vol. 15, Number 1, p. 1905). Zenodo. https://doi.org/10.5281/zenodo.10511579

Code was run in jupyter notebooks. You find them in the directory /notebooks.

• fit_model_1.ipynb : Fits the main model, and computes all analysis plots and tables as seen in the paper.
• model_comp.ipynb : Fits the main model and a null model and uses LOO for model comparison.
• recovery.ipynb : Performs sensitivity analysis across a range of parameters.

/src contains the PyMC models in model.py and other main functions imported in notebooks.

I use the pacakge manager uv to setup the virtual environment. If you do not already have uv installed we refer to [https://docs.astral.sh/uv/getting-started/installation/] For example try, curl -LsSf https://astral.sh/uv/install.sh | sh or brew install uv.

You can then run,

uv sync

and activate .venv

source .venv/bin/activate