EFA vs CFA, saving factor scores
Hi all,
I'm running a SEM model with multiple variables, some observed and some latent. I want to obtain factor scores for the latent variables, so I did a separate EFA (1-factor, 3 indicators) and checked the 'Add FA scores to data' box. But now I have 2 questions (for context, I'm not a statistician):
1) I don't really understand how the calculation of this variable works. Does it transform the raw indicator values with (e.g.) factor loadings, controlled for something else?
2) I also did a CFA (1-factor, same 3 indicators) and saved the factor score, but saw these scores are slightly different. The factor loadings differ as well, so I'm guessing it's got to do with that. But now the question is; which of the factor scores should I use, from the EFA or the CFA?
Hope you can help me with my questions, thanks in advance!
Kind regards,
Lars
Comments
I'll attend our expert to this!
EJ
Hi Lars,
it depends on your research question if you should conduct an exploratory or a confirmatory factor analysis. Depending on that question you should run one of the analyses. It also makes sense that the values from both are slightly different, given the estimation procedure is different.
Regarding the factor scores: Simplified, in factor analysis a participants's score is split into factor loading(s) multiplied by the factor score plus a residuum. So the factor scores are just that, the scores in this decomposition for all observations.
Julius