Resolving Dimensionality in a Child Assessment Tool: An Application of the Multilevel Bifactor Model

• In Event: Methodology and Application in Structural Equation Modeling

Abstract

Clustering of items into testlets and test-takers into groups is fairly common in assessment data. These design features threaten the validity of results and inferences generated from factor analysis, a method frequently employed to assess test dimensionality. In this paper, we illustrate the application of the multilevel bi-factor model as a tool to address these features in examining test dimensionality. The vehicle for our demonstration is Child Observation Record Advantage1.5 (COR-Adv1.5). Results from comparing several models show that the multilevel bifactor model, with one general and three specific factors at the within level and one general and one specific factor at the between level, provides the best fit for the COR-Adv1.5 data. Implications of the identified latent structure are discussed.

Authors

• Hope O. Akaeze, Michigan State University

• Frank Lawrence, Michigan State University

• Jamie Wu, Michigan State University