Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session C: Friday, August 15, 2:00 – 5:00 pm, de Brug & E‑Hall

A hierarchical multivariate copula-based framework for cognitive modeling

Jesper Fischer Ehmsen¹, Arthur Courtin¹, Francesca Fardo¹; ¹Aarhus University

Presenter: Jesper Fischer Ehmsen

Computational cognitive models provide an approach to understanding behavior and cognition by formalizing latent parameters underlying decision-making and learning. Many existing models take a univariate approach, analyzing single measures in isolation, while others incorporate multiple measures but impose specific process assumptions that constrain how these measures relate i.e. drift diffusion models. Here, we introduce a hierarchical multivariate modeling framework that uses copulas to flexibly combine independent likelihood functions, enabling joint modeling of multiple measures without imposing restrictive assumptions. Through simulations and empirical applications, we assess the reliability, discriminability, and advantages of copula-based modeling (CBM). Model validation via simulation-based calibration, model recovery, and sensitivity analyses demonstrate that CBM is computationally robust and accurately recovers latent parameters and their uncertainty. When applied to psychophysical and probabilistic learning tasks, CBM can be empirically distinguished from DDMs, even with limited data. We show that this framework enables efficient use of available data by integrating multiple sources of information, while enhancing model accuracy and efficiency of parameter estimation.

Topic Area: Methods & Computational Tools

Extended Abstract: Full Text PDF