A comparative evaluation of factor- and component-based structural equation modelling approaches under (in)correct construct representations

This study examines how choosing a factor or a component to represent a construct affects the performance of structural equation modeling methods. It clarifies population component models and compares maximum likelihood estimation, factor score regression, GSCA, and partial least squares path modeling across experimental conditions. The findings support matching the estimation approach to the assumed construct representation. In the conditions studied, component methods were generally less sensitive to an incorrect representation, and GSCA performed better than partial least squares path modeling.