Robust Bayesian hypothesis testing with the hierarchical EZ-DDM

The EZ-diffusion model (EZ-DDM) uses a method of moments to provide closed-form estimators for the drift-diffusion parameters from summary statistics. In previous work, we showed that using the sampling distributions of these statistics enables the implementation of an hierarchical extension of the EZ-DDM, supporting scalable Bayesian inference in cognitive psychometrics applications. However, the summary statistics used in EZ-DDM implementations (mean and variance) are sensitive to contaminant data points, limiting its utility in real-world applications with contaminated data. To address this, we develop and test a variation on the EZ-DDM in which we replace the summary statistics with robust alternatives: we substitute mean RT with median RT and RT variance with an estimate derived from the interquartile range. We illustrate the effectiveness of this substitution in a simulation study using a within-subject t test design across varying sample sizes and effect sizes. The robust variant matched the diagnostic accuracy of the original EZ-DDM for uncontaminated data but remained stable under contamination, unlike the standard model. This extension preserves efficiency while adding robustness in real-world applications. We recommend the use of the robust EZ-DDM in practical applications.