A Generalized Speed-Accuracy Response Model for Dichotomous Items.

Area of Science:

• Psychometrics
• Cognitive modeling
• Statistical modeling

Background:

• The speed-accuracy response model (SARM) is used to model scores incorporating response speed and accuracy.
• Existing models may not fully capture the nuances of speed-accuracy trade-offs in cognitive tasks.

Purpose of the Study:

• To propose and evaluate a generalized two-parameter speed-accuracy response model (SARM).
• To develop an expectation-maximization (EM) algorithm for parameter estimation and standard error calculation.
• To provide methods for assessing model fit using generalized residuals and saddlepoint approximations.

Main Methods:

• Generalization of the one-parameter SARM to a two-parameter model, analogous to Rasch to Birnbaum models in IRT.
• Development of an expectation-maximization (EM) algorithm for parameter estimation.
• Implementation of generalized residuals and saddlepoint approximations for model fit assessment.

Main Results:

• A simulation study demonstrated good parameter recovery and acceptable Type I error rates for the proposed methods.
• Application to two real data sets showed that the two-parameter SARM provided a better fit than the one-parameter SARM.
• The developed EM algorithm and model fit assessment tools were effective in analyzing the data.

Conclusions:

• The proposed two-parameter SARM represents a valuable extension of existing speed-accuracy modeling techniques.
• The developed estimation and model fit procedures are robust and applicable to real-world cognitive data.
• This enhanced model offers a more nuanced understanding of the speed-accuracy trade-off in psychological and cognitive research.