A Practical Introduction to Hierarchical Bayesian Drift Diffusion Modeling

Google map

Cognitive models are useful methodological tools for generating insights about latent cognitive processes underlying observable decision-making behavior. The drift diffusion model (DDM; Ratcliff, 1978; Ratcliff and McKoon, 2008) is a prominent framework originally developed to characterize rapid, forced-choice decisions between two response options. By jointly modeling response time and response accuracy, the DDM provides interpretable parameters that can be mapped onto theoretically relevant cognitive mechanisms. The distributional properties of response time data are better represented with this approach, and it accounts for well-studied influences such as the speed-accuracy tradeoff.

This talk provides a practical introduction to fitting drift diffusion models within a hierarchical Bayesian framework. First, key conceptual motivations and modeling considerations are introduced. Next, an overview of the mathematical formulation of the DDM and its core parameters is provided. Finally, we walk through a step-by-step illustrative example in R and JAGS. The demonstration uses data from a within-person, high-frequency ambulatory factorial experiment in which participants completed multiple versions of the Color Shapes task. The example includes task manipulations designed to influence specific DDM parameters, and concludes with an applied model estimation and interpretation.