Interval Estimation of Interclass Correlation Coefficients in Multilevel Item Response Models

• In Event: 66.052 - Estimation Accuracy and Conditions for Complex Multilevel Models

Abstract

The intraclass correlation coefficient (ICC) is an important parameter in multilevel models that address cluster randomized trails. Compared to the point estimate, the confidence interval of ICC has been less explored in multilevel latent variable models. We identify four existing ICC interval estimation methods and propose an alternative method, the multiple imputation based characterization of confidence interval for ICC in multilevel item response models. The performance of these five methods are compared via a Monte Carlo simulation study. The preliminary results of the simulation are presented, and an empirical data illustration will be also provided.

Authors

• Xiaying Zheng, University of Maryland - College Park

• JI SEUNG YANG