A mixture hierarchical model for response times and response accuracy.

Area of Science:

• Educational Measurement
• Psychometrics
• Data Analysis

Background:

• Examinees exhibit diverse test-taking behaviors, including solution and rapid guessing.
• Ignoring these behaviors can violate the local independence assumption, leading to biased parameter estimates in testing.
• Understanding these behaviors is crucial for accurate assessment of both items and individuals.

Purpose of the Study:

• To propose a novel mixture hierarchical model to differentiate between solution behavior and rapid guessing behavior.
• To accurately estimate item and person parameters by accounting for distinct test-taking strategies.
• To enhance the identification of specific examinee behaviors during test administration.

Main Methods:

• Development of a mixture hierarchical model incorporating item responses and response time patterns.
• Utilizing a Monte Carlo expectation maximization algorithm for model calibration.
• Comparison of the proposed model against a non-mixture model using simulation studies and real test data.

Main Results:

• The proposed mixture model provides more accurate item and person parameter estimates compared to non-mixture models when distinct behaviors are present.
• The model effectively distinguishes between solution behavior and rapid guessing behavior in examinees.
• The model demonstrates superior fit to real, high-stakes test data.

Conclusions:

• The mixture hierarchical model successfully accounts for different test-taking behaviors, improving estimation accuracy.
• This approach enhances the ability to identify underlying examinee behaviors, leading to more reliable assessment outcomes.
• The model offers a significant advancement in psychometric analysis for educational and psychological testing.