A Multilevel Approach to Identifying Criterion-Related Profile Patterns

• In Event: Modeling Frameworks for Measurement Issues

Abstract

Profile analysis is a technique that is used to analyze assessments containing strands or subtests. Multilevel profile analysis, assumes there are multiple levels inherent within the data. In this study, we considered and described mathematics profiles that have two inherent levels: An individual and a school level. In order to test whether the school profiles existed and accounted for variability beyond individual profiles, three nested regression models were examined. The results showed the model with random effects for strands at the school level had the best fit for describing the underlying profiles. Our results demonstrated that when a profile was multilevel, it was important for researchers to adopt an appropriate approach that can adequately examine latent profiles at multiple levels.

Authors

• Yu-Feng Chang, University of Minnesota

• Christopher David Desjardins, University of Minnesota

• Chi-Keung Chan, Hong Kong Shue Yan University