Recovery Accuracy of Measurement Model and Structural Coefficients of Extended Bifactor-(S-1) and (S·I-1) Models

Abstract

Bifactor-(S-1) and (S·I-1) models represent general factor measurement models intended to overcome anomalous results associated with bifactor models. Research has centered on model utility using real data, with little consideration of conditions associated with parameter recovery accuracy. Thus, there is a lack of empirical evidence regarding performance of these models under known conditions. This simulation study investigated the parameter recovery of bifactor-(S-1) and (S·I-1) models under varying conditions based on intelligence and psychological test data, including the regression coefficient between the primary dimension to an external criterion. Furthermore, regression coefficient recovery accuracy was compared to results based on a unidimensional model to examine consequences associated with ignoring an instrument’s multidimensional structure.

Keywords:

• Bifactor models
• Monte Carlo
• parameter recovery
• prediction