Title: Recent Developments in Factor Score Regression

Authors: Ines Devlieger, Yves Rosseel

Affiliation: Ghent University

Abstract:
In recent years, factor score regression (FSR) has been proposed as an
alternative method to estimate structural equation models (SEM). The FSR
method can handle misspecifications and small sample sizes better than the
standard maximum likelihood estimator. In FSR, factor scores are calculated as
proxies for the latent variables, which are then used in a regression analysis
or path analysis. In combination with a correction method developed by Croon
(2002), FSR provides unbiased estimates of the structural parameters of the
model. Recently, the FSR method has been implemented in the R package lavaan.
This implementation includes several new developments: analytical standard
errors, local and global  t measures, model comparison tests, and support for
multilevel data. In this presentation, I will illustrate lavaan's FSR
capabilities using an example.