Bias and Efficiency of Estimation in Multilevel Confirmatory Factor Models

• In Event: Issues of Measurement Invariance in Mixture Models for Educational Data

Abstract

Multilevel measurement models are frequently applied to help answer questions when data arise from hierarchically structured multivariate data. We evaluated the efficiency of estimating model parameter among three estimation methods (MLR, ULSMV, WLSMV) for multilevel factor models. For nearly all sample size conditions, WLSMV resulted in more efficient parameter estimates, but MLR resulted in more efficient standard errors. Additionally, we found that all three estimation methods may overestimate the level-2 factor covariance matrix, but, on average, WLSMV resulted in the least amount of bias in parameter estimates. Our talk will include additional nuances of how parameters and standard errors are influenced by sample size and ICC conditions.

Authors

• R. Noah Padgett, Baylor University

• Grant B. Morgan, Baylor University