Concurrent Use of Response Time and Response Accuracy for Detecting Examinees with Item Preknowledge

Seo Young Lee,James A. Wollack

DOI: https://doi.org/10.4324/9781351064781-12

2020-06-03

Abstract:It has been well established that item response times can provide useful insight into examinees' test taking behaviors because observed response time patterns reflect the underlying nature of the problem-solving process. This chapter describes a new mixture model that focuses on detection of examinees with item pre knowledge. It begins by describing the model and provides an example of use of the model with operational examination data. The real-data-based simulation studies using the mixture Rasch-Lognormal response time (MixRL) model demonstrated that the mixture model approach may be viable for the detection of item pre knowledge when the data contain perceivable evidence of item pre knowledge. In the comparison with other mixture models (i.e., mixture Rasch model, mixture Ln-RT model, and Meyer's model), the MixRL model outperformed all other mixture models for the detection of preknowledge in most simulated conditions. These results imply that the MixRL model is a promising approach for the detection of item preknowledge.