The exploratory factor analysis (EFA) model revisited

A. Satorra

EFA is perhaps the most extensively used form of multivariate analysis. It is used in standard routines for multivariate analysis of SPSS, Stata, SAS, etc, and also as a particular class of structural equation models (SEMs). Practitiones typically begins with a sequence of likelihood ratio goodness of fit tests to help determine the number of factors. In recent work, Jennrich and Satorra (2015) give an example were slight deviation from normality can make the usual LRT fail seriously and propose a new test statistic that does not use the normality assumption. The new test statistic can be used also in conjunction with the PFA algorithm for estimation. Here we review the classical and SEM approaches to EFA and offer a comparative analysis of the different methods for goodness of fit testing. Details of the practical implementations of these methods with current software are also discussed.

Palabras clave: Factor analysis, Structural Equation models, Goodness of Fit testing, normal theory, distribution free methods

M05.2 Grupo de Análisis Multivariante y Clasificación III
6 de septiembre de 2016  11:00
0.09 - Aula de proyectos 2

J. Fortiana Gregori, A. Esteve Gómez

J. L. Vicente Villardón

J. F. Vera Vera