A Psychometric Perspective on the Associations between Response Accuracy and Response Time Residuals

Area of Science:

• Psychometrics
• Educational Measurement
• Cognitive Psychology

Background:

• Empirical studies show statistical dependencies between response accuracy residuals (RARs) and response time residuals (RTRs).
• Existing joint models for response time and accuracy do not fully capture these observed associations.
• Recent generalizations of the D-diffusion IRT model highlight person-by-item interactions.

Purpose of the Study:

• To propose an alternative psychometric perspective on the relationship between RARs and RTRs.
• To demonstrate how this perspective aligns with and explains recent empirical findings.
• To explore the applied and interpretational implications of this new viewpoint.

Main Methods:

• Development of a psychometric model emphasizing exogenous influence of RARs on RTRs.
• Utilizing both empirical data and simulation studies for validation.
• Comparison with existing joint response time/accuracy models and D-diffusion IRT generalizations.

Main Results:

• The proposed psychometric perspective successfully reproduces empirical findings on RAR-RTR associations.
• Evidence suggests that item parameter heterogeneity related to response times may be a psychometric artifact.
• The influence of RARs on RTRs needs consideration when using RTRs as indicators of latent proficiency.

Conclusions:

• The alternative psychometric perspective provides a theoretically justifiable explanation for RAR-RTR dependencies.
• Findings suggest re-evaluating the interpretation of item parameter heterogeneity and response time data.
• The choice of perspective may depend on the specific test setting and theoretical considerations.