I want to use structural equations for modeling. From what I understand, the PLS approach is a predictive model based on components, we build components from observed data while the SEM ( traditional ) is a confirmatory model based on covariances, we never calculate scores for latent variables, we only estimate covariances in order to verify the validity of the model built.
In my case I have a set of latent variables (Big five, optimism, narcissism) for which I have the possibility to calculate the score based on the items.
So if I understand correctly I have to use the PLS approach and not the classical SEM?
Many thanks for your help.
Dear Achbah Rachid
You need to use covariance based SEM, if you need to test a theory. If you only need to find out the score or want to predict your model then variance based SEM/PLS is better choice.
This is the first time I'm going to use this method, so I have some questions. Indeed, if I understand correctly, I will use the SEM method based on covariance. Since I am testing a theory that is well developed.
In my model, I have independent variables (Big five, optimism and CEO narcissism). Using the data collected, I can calculate a score for each independent variable. My question is this: Should I put the scores of my independent variables calculated from the items (example of NPI 16 for the level of Narcissisme) ? Or will I create latent variables and test the validity of the measure.
So I am confused between using the calculated scores for each variable and using latent variables.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling