Improving parameter recovery for conflict drift-diffusion models
- PDF / 4,799,993 Bytes
- 19 Pages / 595.276 x 790.866 pts Page_size
- 71 Downloads / 206 Views
Improving parameter recovery for conflict drift-diffusion models Ronald Hübner 1 & Thomas Pelzer 1
# The Author(s) 2020
Abstract Several drift-diffusion models have been developed to account for the performance in conflict tasks. Although a common characteristic of these models is that the drift rate changes within a trial, their architecture is rather different. Comparative studies usually examine which model fits the data best. However, a good fit does not guarantee good parameter recovery, which is a necessary condition for a valid interpretation of any fit. A recent simulation study revealed that recovery performance varies largely between models and individual parameters. Moreover, recovery was generally not very impressive. Therefore, the aim of the present study was to introduce and test an improved fit procedure. It is based on a grid search for determining the initial parameter values and on a specific criterion for assessing the goodness of fit. Simulations show that not only the fit performance but also parameter recovery improved substantially by applying this procedure, compared to the standard one. The improvement was largest for the most complex model. Keywords Drift-diffusion models . Parameter recovery . Model-fit procedure . Grid-search method
Introduction The ability to act in a goal-oriented manner is an essential characteristic of human performance. To investigate involved mental processes, several so-called conflict paradigms have been developed, such as the Stroop task (Steinhauser & Hübner, 2009; Stroop, 1935), the Eriksen flanker task (Eriksen & Eriksen, 1974), and the Simon task (Hübner & Mishra, 2013; Proctor, 2011; Simon, 1969), where irrelevant stimulus features produce response conflicts that are reflected by congruency effects. Recently, conflict DDMs (drift diffusion models) have been proposed that, based on response-time (RT) distributions and accuracy data, model the dynamics of the performance in conflict tasks. The first of these models was the Dual-Stage Two-Phase (DSTP) model (Hübner, Steinhauser, & Lehle, 2010), followed by the Shrinking Spotlight (SSP) model (White, Ratcliff, & Starns, 2011). Both models were first applied to the flanker task. Later, the Diffusion Model for Conflict (DMC) tasks (Ulrich, Schröter, Leuthold, & Birngruber, 2015) has been proposed, which has also been applied to Simon-task data. If interpreted
* Ronald Hübner [email protected] 1
Department of Psychology, Universität Konstanz, D-78457 Konstanz, Germany
accordingly, however, the DSTP model can be used as well to model Simon-task data (Hübner & Töbel, 2019). In studies, in which conflict tasks are modeled, one or several of the considered models are fitted to experimental data in the same way as common DDMs, and much effort is usually spent to obtain good fits. However, although a good fit is important for modeling, it does not guarantee that the obtained parameters are valid (Pitt & Myung, 2002; Roberts & Pashler, 2000). Indeed, it is possible that a model fits data satisfactoril
Data Loading...
