Title:Bayesian Semiparametric Longitudinal Drift-Diffusion Mixed Models for Tone Learning in Adults

Abstract:Understanding how the adult humans learn to to categorize novel auditory categories is an important problem in auditory behavioral neuroscience. Drift diffusion models are popular, neurobiologically relevant approaches to assess the mechanisms underlying speech learning. Motivated by these problems, we develop a novel inverse-Gaussian drift-diffusion mixed model for multi-alternative decision making processes in longitudinal settings. Our methodology builds on a novel Bayesian semiparametric framework for longitudinal data in the presence of a categorical covariate that allows automated assessment of the predictor's local time-varying influences. We design a Markov chain Monte Carlo algorithm for posterior computation. We evaluate the method's empirical performances through synthetic experiments. Applied to a speech category learning data set, the method provides novel insights into the underlying mechanisms.