Exploratory structural equation modelling : a more suitable method for the analysis of latent structure in psychodiagnostics?
Abstract
Confirmatory factor analysis (CFA) may be too strict method to analyze the latent structure of psychodiagnostic methods with a multifactor structure and multiple items. One of the reasons is the forced fixation of cross-loadings and covariance of residuals of individual items, which significantly reduces the model fit.
Attempts to improve the model using modification indices leads in some cases to models that are too complicated and difficult to interpret. Tihomir Asparouhov and Bengt Muthén (2009) developed an exploratory structural equation modeling (ESEM) method especially usefull for these cases.
One of the main benefits of ESEM is a more flexible approach to the model in the spirit of exploratory factor analysis (EFA) enabled by the tolerance of small covariates in the model. However, ESEM, unlike the EFA, is primarily a confirmatory approach.
The aim of the paper is to briefly introduce ESEM and start discussion about the benefits, limitations and possibilities of using this approach as an alternative to more commonly used methods.