Published on

HSSM - Hierarchical Sequential Sampling Modeling

  • avatar

HSSM - Hierarchical Sequential Sampling Modeling

📦 HSSM is a Python toolbox that combines state-of-the-art likelihood approximation methods with the wider ecosystem of probabilistic programming languages. It enables flexible hierarchical model building and inference through modern MCMC samplers. HSSM is designed to estimate the impact of neural and other trial-by-trial covariates through parameter-wise mixed-effects models for various cognitive process models.

HSSM is a project developed in collaboration with the Center for Computation and Visualization and the Center for Computational Brain Science within the Carney Institute at Brown University.

How does it work?

HSSM allows for approximate hierarchical Bayesian inference using different likelihood approximators. It provides native support for hierarchical mixed-regression, enabling the estimation of the impact of neural and other covariates on a trial-by-trial basis. HSSM is built on top of PyMC and leverages the Python Bayesian ecosystem. It also incorporates Bambi's intuitive lmer-like regression parameter specification for within- and between-subject effects. The toolbox includes native support for plotting and other convenience functions to aid the Bayesian workflow through the integration of ArviZ. Additionally, HSSM utilizes the ONNX format for the translation of differentiable likelihood approximators across backends.

Benefits and Use Cases

HSSM offers several benefits and can be used in various scenarios:

🔬 Cognitive Modeling: HSSM provides a powerful framework for building and inferring hierarchical cognitive process models. It allows researchers to estimate the impact of neural and trial-by-trial covariates on cognitive processes.

📊 Mixed-Effects Modeling: HSSM supports mixed-effects models, enabling the estimation of both within-subject and between-subject effects. This is particularly useful when analyzing data with nested structures or when accounting for individual differences.

📈 Flexible Inference: HSSM offers a range of likelihood approximators, allowing for flexible and approximate Bayesian inference. Researchers can choose the method that best suits their modeling needs.

🌐 Integration with Python Ecosystem: HSSM is built on top of PyMC and leverages the wider Python Bayesian ecosystem. This integration provides access to a rich set of tools and libraries for Bayesian modeling and analysis.

Future Directions

The HSSM project is actively developed and maintained by the team at Brown University. The developers are continuously working on improving the toolbox and adding new features. Some of the future directions for HSSM include:

🔍 New Models and Likelihoods: The HSSM community is encouraged to contribute new models and likelihoods to expand the toolbox's capabilities. Users can add novel models with corresponding likelihoods to address specific research questions.

📚 Documentation and Tutorials: The HSSM project aims to provide comprehensive documentation and tutorials to help users get started and make the most of the toolbox. This includes examples, step-by-step guides, and best practices for different modeling scenarios.

🌟 Community Support: The HSSM project is committed to fostering a supportive and active community. The developers encourage users to ask questions, report bugs, and contribute to discussions on the GitHub repository.


HSSM is a powerful Python toolbox for hierarchical sequential sampling modeling. It enables flexible hierarchical model building and inference using state-of-the-art likelihood approximation methods. With its support for mixed-effects models and integration with the Python Bayesian ecosystem, HSSM is a valuable tool for researchers in the field of cognitive modeling. The project is actively developed and welcomes contributions from the community. To learn more about HSSM and get started, visit the official documentation.

🔗 HSSM GitHub Repository