pyhddmjags

(Repository version 0.7.1)

Repository for example Hierarchical Drift Diffusion Model (HDDM) code using JAGS in Python

Authors: Michael D. Nunez from the Psychological Methods group at the University of Amsterdam

Possible citation

Nunez, M. D., Fernandez, K., Srinivasan, R., & Vandekerckhove, J. (2024). A tutorial on fitting joint models of M/EEG and behavior to understand cognition. Behavior Research Methods. https://doi.org/10.3758/s13428-023-02331-x

Prerequisites

Python 3 and Scientific Python libraries

For these next install steps in Ubuntu, see jags_wiener_ubuntu.md in this repository.

MCMC Sampling Program: JAGS

Program: JAGS Wiener module

Python Repository: pyjags, can use pip:

pip install pyjags

See also yaml/pyjags.yml

Optional Stan + PyStan

The newest version of PyStan was PyStan 3, but I didn't find PyStan 3 as easy to use with custom diagnostic and plotting scripts as PyStan 2.

Here are the docs for PyStan 2

See also yaml/pystan.yml

Downloading

The repository can be cloned with git clone https://github.com/mdnunez/pyhddmjags.git

The repository can also be downloaded via the Code -> Download zip buttons above on this Github page.

Getting started

At the moment each script can be run individually to simulate from hierarchical drift-diffusion models (HDDMs) and then find and recover parameter estimates from those models. The most simple HDDM lives in nolapse_test.py. See other scripts: recovery_test.py, blocked_exp_conds.py, and regression_test.py. These scripts provide useful examples for using JAGS with pyjags, the JAGS Wiener module, mixture modeling in JAGS, and Bayesian diagnostics in Python.

The script nolapse_test_pystan.py contains pystan and Stan code to find and recover parameters from the exact same HDDM written in JAGS within nolapse_test.py. Note that this script will not work with the newest versions of pystan (pystan 3). One solution is to create a Python environment using conda:

conda create -n mcmc python=3 numpy scipy pystan==2.19
conda activate mcmc

Note that you may need or want to create a separate environment for pyjags using conda:

conda create -n jags python=3.8 numpy
conda activate jags
pip install pyjags

License

pyhddmjags is licensed under the GNU General Public License v3.0 and written by Michael D. Nunez from the Psychological Methods group at the University of Amsterdam.

Usage examples

Nunez, M. D., Gosai, A., Vandekerckhove, J., & Srinivasan, R. (2019). The latency of a visual evoked potential tracks the onset of decision making. NeuroImage. doi: 10.1016/j.neuroimage.2019.04.052 (See repository encodingN200 associated with this paper for specific pyjags examples.)

Lui, K. K., Nunez, M. D., Cassidy, J. M., Vandekerckhove, J., Cramer, S. C., & Srinivasan, R. (2021). Timing of readiness potentials reflect a decision-making process in the human brain. Computational Brain & Behavior. (See repository RPDecision associated with this paper for a specific HDDM in JAGS using MATLAB.)

Nunez, M. D., Vandekerckhove, J., & Srinivasan, R. (2017). How attention influences perceptual decision making: Single-trial EEG correlates of drift-diffusion model parameters. Journal of Mathematical Psychology, 76, 117-130. (See repository mcntoolbox associated with this paper for a specific HDDM in JAGS using MATLAB.)

Nunez, M. D., Srinivasan, R., & Vandekerckhove, J. (2015). Individual differences in attention influence perceptual decision making. Frontiers in Psychology, 8. (This paper fit HDDMs in JAGS like that in regression_test_py3.py using MATLAB and Trinity.)