A flexible framework for simulating and fitting generalized drift-diffusion models.

Dr. John Murray, Assistant Professor, Department of Psychiatry, Division of Neurocognition, Neurocomputation, and Neurogenetics (N3), Yale School of Medicine

Watch:

Presentation Slides:

The drift-diffusion model (DDM) is a well-defined computational model that assumes that in a 2-option forced choice task, the subject is accumulating evidence for one or other of the alternatives at each time step, and integrating that evidence until a decision threshold is reached. Individual differences in the process of evidence accumulation can be quantified using a small set of parameters. Basic DDMs are well-established in psychology and neuroscience. We introduce a computational framework for simulating and fitting "generalized drift-diffusion models" (GDDMs) for choice and response-time data, and an associated software package PyDDM.

The primary advantage of our GDDM approach is that it can flexibly instantiate time-varying inputs and a variety of decision processes that researchers are interested in (e.g., urgency, leak, time-dependent biases), whereas basic DDMs cannot. The challenge to use of GDDMs in the field is the computational cost of simulation and fitting, for which PyDDM provides a solution.

GDDMs have been proposed and studied, but their application to experimental data analysis had been hindered by technical challenges, which is why we developed PyDDM. Our paper was published this year in eLife: Shinn M*, Lam NH*, Murray JD (2020) A flexible framework for simulating and fitting generalized drift diffusion models. eLife 9:e56938.

The PyDDM software package is implemented in Python/NumPy, so some ability to work with Python/NumPy is required. The PyDDM software package is free and open-source, written in Python/NumPy. PyDDM can be download from the Murray Lab Github page.

The package can be applied to any dataset that includes choices and response times in a two-alternative forced-choice task paradigm.
Tutorials, documentation, and FAQ for PyDDM can be found online.

Publication: https://elifesciences.org/articles/56938
Code: https://github.com/murraylab/PyDDM
Documentation: https://pyddm.readthedocs.io
PyDDM mailing list sign-up.