Modelos de variables latentes para la investigación conductual

Abstract

An introduction to modelling latent variable as an analytic strategy for data in behavioral research is presented. This strategy allows to model and manipulate constructs, in contrast to other analytical aproaches that only deal with observed variables. Procedures such as exploratory factor analysis, confirmatory factor analysis and structural equations models are detailed as means to model latent variables. It is emphazised that the use of structural equations models allows the "construction" of latent variables, the modelling of relations between these constructs, other observed variables and emergent variables. Some applications of these analytic systems are presented, such as indirect effects estimation, causal direction, stability, internal consistency, convergent validity, discriminant validity, and predictive validity of behavioral measures, as well as the especification and estimation of treatment effects and group comparisons. The usefulness of these applications to scientific behavioral research is discussed.