Identifying Effortful Individuals With Mixture Modeling Response Accuracy and Response Time Simultaneously to Improve...

Area of Science:

• Psychometrics
• Educational Measurement
• Data Analysis

Background:

• Non-effortful responding in testing can significantly impair the accuracy of model calibration and latent trait inferences.
• Identifying and accounting for non-effortful responses is crucial for valid assessment outcomes.

Purpose of the Study:

• To introduce and evaluate novel mixture models for differentiating between non-effortful and effortful individuals.
• To enhance item parameter estimation by focusing on effortful responses.
• To compare the performance of new mixture models against traditional methods like the Response Time Mixture Model (TMM) and Normative Threshold 10 (NT10).

Main Methods:

• Development of two mixture model approaches incorporating both response accuracy and response time.
• Comparison of these mixture models with TMM and NT10 methods.
• Simulation studies across four scenarios to assess item parameter recovery and classification accuracy.

Main Results:

• Mixture methods and TMM effectively reduce bias in item parameter estimates caused by non-effortful responding.
• Mixture methods demonstrate superior performance, especially when non-effort is severe or response times deviate from lognormal distribution.
• Both methods showed improvements in classification accuracy compared to baseline.

Conclusions:

• Mixture models offer a robust approach to address the challenges posed by non-effortful responding in assessments.
• These models provide more accurate latent trait inferences and item parameter estimates.
• The proposed methods are particularly beneficial in complex response time distributions or high levels of non-effort.