A Latent Markov Model for Interaction Log Analytics

• In Event: Refining Computerized Testing: Precision, Validity, and Innovation in Educational Measurement

Abstract

The study proposes a refinement of the latent Markov model (LMM) for interaction log data from computer-interactive assessments. We extend the current development in LMM to support transition analysis of multimodal interaction sequence data and enable inference on the measurement parameters at the individual event level. Numerical experimentations are conducted to verify the reliability and applicability of the new framework. The experimental observations suggest that the new framework achieves adequate inferential reliability and provides credible decoding outcomes while appropriately accounting for items’ measurement effects. Based on the empirical findings, we discuss the potential of the new LMM framework for serving large-scale interaction log analytics.

Author

• Hyeon-Ah Kang, University of Texas at Austin