Mitigating Retest Effects in Cognitive Ecological Momentary Assessment Data - A Tutorial in PyMC, Stan and JAGS

This tutorial provides a comprehensive, step-by-step guide for implementing the Bayesian Double Exponential Model (BDEM) in PyMC, Stan, and JAGS. The BDEM is a statistical framework designed to disentangle retest (practice) effects from true longitudinal cognitive change in high-frequency Ecological Momentary Assessment (EMA) and measurement burst designs. We highlight the unique features of the three approaches in terms of accessibility, flexibility, sampling algorithms, and computational time, focusing on their practical implications for implementing complex hierarchical models. A simulation study was conducted to evaluate the performance of these software programs across both small and large sample sizes. Results indicated that while all three programs successfully recover most model parameters, Stan yielded the highest sampling quality, whereas JAGS offered the lowest computational time under the same MCMC settings (e.g., number of total iterations). Learning rates remained the most challenging features to estimate, particularly in small samples. This tutorial equips researchers with the necessary tools to apply advanced multilevel cognitive modeling to high-frequency cognitive assessments in measurement burst designs.