Gaze-dependent evidence accumulation predicts multi-alternative risky choice behaviour
This repository contains all raw data, preprocessing, analysis and visualization code used in the paper:
- Molter, F., Thomas, A. W., Huettel, S. A., Heekeren, H. R., & Mohr, P. N. C. (2022). Gaze-dependent evidence accumulation predicts multi-alternative risky choice behaviour. PLOS Computational Biology, 18(7), e1010283. https://doi.org/10.1371/journal.pcbi.1010283
All analyses* are written in Python.
Dependencies and package versions used are listed in the environment.yml file, which you can use to reproduce the computing environment (e.g., using the Anaconda Python distribution, see instructions here). The main dependencies of this project are the standard Python data stack (numpy, pandas, scipy), PyMC3, Theano, bambi and pyyaml for statistical analyses, matplotlib, seaborn and python-ternary for visualization.
All Python analyses can be run in sequence by calling the run_all_analyses.sh script. This script calls the individual analysis scripts in the src folder.
Note that the project involves fitting of many (~200) behavioural model variants, which can take days to weeks, depending on the machine. By default, model fitting results are not overwritten, but read from the repository. If you plan to reproduce the model-fitting, you can adapt the overwrite and ncores command line arguments for the 3-1_behavioural-modeling_fitting.py and 4-1_switchboard_fitting.py scripts.
Statistical analyses use sampling-based Bayesian estimation methods and can yield slightly different results each run, even if random seeds are set, due to powers that I don't understand.
Contact
Questions or comments should be addressed at felixmolter@gmail.com.
* The BMS script 3-3_run_bms.m must be run in MATLAB manually. Alternatively, the code includes a basic Python implementation of a Bayesian Model Selection procedure in src/analysis/bms.py that can be used to compute basic exceedance probabilities in Python.