Dear all, I have a data set with 7 variables. I have run a EFA and I got three factors. I want to obtain a numerical value for each factor computed from the variables that form it but I am not sure how to proceed.
Let's say I have Factor 1, which is influenced by variable 1 (loading 0.80) and variable 2 (loading -0.76).
If I want to compute a numerical value for Factor 1 based on those two variables is it enough to:
Factor 1=Variable 1*0.80 - variable 2*0.76?
Or maybe I should run a SEM?
Thank you very much!
Hi Clara,
there are different options for creating a composite but apart from that, they will not result in a latent variable but in a sum of error-containing indicators. Hence, going "full-SEM" is surely the better solution.
Whether you compute weighted composites or a simple sum of the indicators makes not so much of a difference but there is some debate about that:
Bobko, P., Roth, P. L., & Buster, M. A. (2007). The usefulness of unit weights in creating composite scores: A literature review, application to content validity, and meta-analysis. Organizational Research Methods, 10(4), 689-709.
With regard to the error contamination (called factor indeterminacy), see
Grice, J. W. (2001). Computing and evaluating factor scores. Psychological methods, 6(4), 430.
HTH
Holger
Holger Steinmetz thank you for your help, I will check those references to see how I proceed.
Hi Clara,
concerning the indicators Holger mentioned: SPSS offers three methos to calculate those indicators (regression=the approach you thought of, Bartlett, Anserson-Rubin). You find these indicators in the menue "factor scores" in "factor analysis".
Christian
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running