An Application of a Structured Mixture Rasch Model to Computer Adaptive Data

• In Event: AERA Poster Session 2
  In Poster Session: Poster Session 1 (San Diego)

Abstract

Mixture Item Response Theory (IRT) is a useful approach to modeling latent heterogeneity in student responses and is typically applied in an exploratory manner. This paper, however, demonstrates the application of a structured confirmatory mixture IRT model to computer adaptive test data by exploring latent classes in early kindergarten geometry performance on Common Core subdomains. The analysis shows that this model is useful in exploring the relationships among item subdomains, but that the test design must be considered when looking at individual student classifications. Results suggest the existence of three latent classes that represent classes of students with strengths in different geometry subdomains. Results have implications for future applications of the model and for differentiation of instruction in early kindergarten.

Author

• Meredith Langi, NWEA