Exploratory structural equation modeling: A new trend of factor analysis

Structural equation modeling (SEM) is a multivariate statistical method to estimate the causal relationship among latent variables commonly employed in education and social sciences. Traditionally, confirmatory factor analysis (CFA) measurement models are used prior to the SEM analysis. Marsh (2007) as well as Marsh, Hau, and Grayson (2005) indicated that many psychology measuring instruments cannot even meet the minimum goodness-of-fit criteria. Asparouhov and Muthén (2009) indicated that SEM using CFA measurement model cannot achieve satisfactory model fit frequently, and yet extensive model modifications are required. Misspecification of zero loadings would lead factor distortion with overestimated factor correlations. Therefore, Asparouhov and Muthén (2009) suggested that SEM with exploratory factor analysis (EFA) measurement model is a better technique to avoid these issues. Instead of confirming the relationships among observed variables and latent variables prior to performing SEM, the EFA-SEM (ESEM) approach with rotation is incorporated within the SEM. This commentary discusses the benefits and restriction of SEM using confirmatory approach and the benefits of SEM by exploratory approach.