A Multivariate Multilevel Approach to the Modeling of Accuracy and Speed of Test Takers

Area of Science:

• Psychometrics
• Statistical Modeling
• Educational Measurement

Background:

• Modern computerized testing readily collects response times alongside accuracy data.
• Understanding the interplay between response speed and accuracy is crucial for comprehensive test analysis.

Purpose of the Study:

• To develop and validate a statistical framework for jointly modeling response accuracy and response times.
• To investigate individual and group-level factors influencing both speed and accuracy in test-taking.

Main Methods:

• Extension of existing models with a multivariate multilevel regression structure.
• Application of a Bayesian approach utilizing Markov chain Monte Carlo (MCMC) computation.
• Development of model-selection criteria including Bayes factor (BF) and Deviance Information Criterion (DIC).

Main Results:

• The proposed model effectively estimates parameters related to speed-accuracy relationships.
• Simulation studies and real-data examples demonstrate the model's utility.
• Covariates can be incorporated to explain variance in speed and accuracy.

Conclusions:

• The developed Bayesian framework offers a powerful tool for analyzing response times and accuracy in computerized tests.
• This approach facilitates a deeper understanding of test-taker behavior and performance variations.
• The model supports novel analyses for educational and psychological measurement research.