Title: More Estimation of Speed-Accuracy Response Models

Author: Peter van Rijn

Affiliation: ETS Global
 
Abstract: 
At Psychoco 2014, I presented some initial ventures into estimating
speed-accuracy response models as presented in a paper by Maris and van der Maas
(2012, Psychometrika). Over the past three years, the software has matured and
now includes marginal maximum likelihood estimation of two types of
speed-accuracy response models (a one-parameter and two-parameter version) using
expectation-maximization, Newton-Raphson, or a combination of both. Integration
is performed using adaptive Gauss-Hermite quadrature. In addition to parameter
estimates, three types of standard errors are provided (regular, Louis, and
sandwich) and two ability estimators are implemented (ML and EAP). Thirdly, two
types of statistics based on generalized residuals are provided to evaluate
model fit. The software will be illustrated with both simulated and real data
examples.