Maximum likelihood estimation of dynamic structural equation models for ICT impact

Abstract

We propose estimation methods for dynamic structural equation models with latent variables that generalize models such as that of M. Skrtic et al. (2003) which considers the impact of Internet-use on research output. The proposed methods are based on the full information maximum likelihood (FIML) principle and have better small sample performance then existing distribution-free methods based on instrumental variables (D. Cziraky, 2003)