I intend to use factor scores as derived from exploratory factor analysis in subsequent multivariate regression analysis, as an explanatory variable.
I've read in multiple books/papers that Structural Equation Modeling is the appropriate method to model relationships AMONG LATENT VARIABLES (using factor scores being incorrect). I haven't, so far, found a negative answer as to whether it is a problem to use factor scores solely as an explanatory variable, my dependent variable being observed.
Does anyone have a clear answer regarding this issue?
Structural equation modeling may be the appropriate method. It tends to be most useful and valid when you have multiple links that you want to identify in a causal chain; when multivariate normality is present; when any missing data are missing completely at random; when N is fairly large; and (I think) when variables are measured without much error. Absent such conditions, exploratory factor analysis scores may be quite useful as regression predictors, assuming the EFA (as well as the regression) is done in a sound, thoughtful way. A lot of people make the mistake of treating EFA as a routinized procedure, as you can read about in the wonderful article Repairing Tom Swift's Electric Factor Analysis Machine. EFA involves many decision points and few iron-clad guidelines for them. 42.2% of all EFA solutions that I run across smack of what I believe to be significant errors in choice of extraction method, number of factors to extract, inclusion/exclusion of variables, or others.