# Whether to use EFA or CFA to predict latent variables scores?

I have a dataframe of individual observations, that I partitioned to create a training (0.7 prop) and a test set (0.3 prop).

I started by running an exploratory factor analysis (EFA) on the training set (with the psych::fa function). Then, based on results, I fitted a CFA on the test set (with lavaan), associating each observed variable with one latent factor based on its maximum loading.

I would like, then, to use these latent variables as new variables for other analyses (using factor analysis as a feature reduction method). I know it's best to do structural equation modelling, as those latent variables include error, but that's not the option I want to pursue in respect to the models I want to fit.

Anyway, should I use the results of the EFA or the CFA to predict the new factors on my initial dataframe?

• It probably won't make much difference. Why not try both? – Jeremy Miles May 16 '18 at 16:34
• Altough in EFA, all variables load on all components (even if some of them load very weakly), as in CFA, the components are created with a subset of variables (those that load maximally on them). I have the feeling that the predictions based on EFA would be more "fine-grained", integrating information of all variables in each component. But at the same time, it sounds a bit strange to do EFA, then CFA, then come back to EFA for predictions... – Dominique Makowski May 17 '18 at 6:31
• The EFA might be more fine grained, but also might contain noise, because there won't be any zero loadings, even if they're zero in the population. – Jeremy Miles May 17 '18 at 22:19
• So if I understand correctly, there is no "gold standart" procedure or particular reasons or cautions beyond those we said for or against these two methods? It seems weird to try both and select the one that works the best... – Dominique Makowski May 18 '18 at 8:18