The seven-parameter diffusion model

• Verfasst von: Henrich, Franziska [VerfasserIn]
• Verfasst von: Hartmann, Raphael [VerfasserIn]
• Verfasst von: Pratz, Valentin [VerfasserIn]
• Verfasst von: Voß, Andreas [VerfasserIn]
• Verfasst von: Klauer, Karl Christoph [VerfasserIn]
• Titel: The seven-parameter diffusion model
• Titelzusatz: an implementation in stan for bayesian analyses
• Verf.angabe: Franziska Henrich, Raphael Hartmann, Valentin Pratz, Andreas Voss, Karl Christoph Klauer
• Jahr: 2024
• Umfang: 15 S.
• Illustrationen: Illustrationen
• Fussnoten: Online veröffentlicht: 28. August 2023 ; Gesehen am 16.10.2023
• Titel Quelle: Enthalten in: Behavior research methods
• Ort Quelle: New York, NY : Springer, 2005
• Jahr Quelle: 2024
• Band/Heft Quelle: 56(2024), 4, Seite 3102–3116
• ISSN Quelle: 1554-3528
• DOI: doi:10.3758/s13428-023-02179-1
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: Bayesian inference
• Sach-SW: Model fitting
• Sach-SW: Ratcliff diffusion model
• Sach-SW: Stan function
• K10plus-PPN: 186573151X

Abstract

Diffusion models have been widely used to obtain information about cognitive processes from the analysis of responses and response-time data in two-alternative forced-choice tasks. We present an implementation of the seven-parameter diffusion model, incorporating inter-trial variabilities in drift rate, non-decision time, and relative starting point, in the probabilistic programming language Stan. Stan is a free, open-source software that gives the user much flexibility in defining model properties such as the choice of priors and the model structure in a Bayesian framework. We explain the implementation of the new function and how it is used in Stan. We then evaluate its performance in a simulation study that addresses both parameter recovery and simulation-based calibration. The recovery study shows generally good recovery of the model parameters in line with previous findings. The simulation-based calibration study validates the Bayesian algorithm as implemented in Stan.