Title: DstarM: An R Package for Analyzing Two Choice Reaction Time Data with the
D*M Method

Authors: Don van den Bergh, Francis Tuerlinckx, Stijn Verdonck

Affiliation: University of Amsterdam

Abstract:
Two-choice reaction time data are gathered frequently in both psychology and
neuroscience. The vast majority of models used to describe such data assume
that observed reaction times are the sum of decision processes and residual
nondecision processes. Commonly, decision processes are modelled with diffusion
models (e.g., the Ratcliff diffusion model or the linear ballistic accumulator
model), while nondecision processes are modelled according to a uniform
distribution. However, the choice for such a uniform nondecision distribution is
not empirically motivated; it is convenient to do so from a modelling
perspective. Moreover, it has been shown that a misspecification of the
nondecision distribution can lead to biased parameter estimates of the decision
model (Ratcliff, 2013). Consequently, conclusions might be biased as well.
Recently, the D*M method (Verdonck & Tuerlinckx, 2016) for estimation diffusion
models has been developed that does not make any distributional assumptions
about the nondecision processes. Therefore parameter estimates of the decision
model are less biased. In this paper, we present the R-package DstarM, which is
a collection of functions that implements the D*M method. We first give a brief
introduction of diffusion models. Next, we explain traditional estimation
methods and the D*M method, followed by a comparison of both methods in a
simulation study. Then, we will give a tutorial on the D*M method by analyzing
an empirical dataset. We conclude with some limitations of the D*M method and
how they can be solved by the design of a study.