I have two layer measurement model (i.e, main construct - subconstruct -measured items) with three main constructs and four to three sub-constructs for each main construct. Again the subconstructs are represented by measured variables.
I run CFA and SEM for the two layer measurement model.
I got different factor loadings(i.e., estimates) for the same measured items( also for sub-constructs) in confirmatory factor analysis and structural equal modelling.
what are the reasons for such differences?
thanks in advance.
Little bit difference can be found usually, but not a big different to be afraid,. And when u confirmed ur items loading in CFA, there is no need to look again in SEM for items loading. In SEM, items loading become useless. Focus on CFA factor loading which require for AVE and CR etc.
Dear Srikanth,
I am surprised to note your finding.Structural equation modeling software is typically used for performing confirmatory factor analysis. CFA is based on the prior literature. Identify the model fit indexes.
Regards,
Prof. Yogesh Pai P
Dear Srikanth,
I agree with Anwar. Usually, the difference should be small. Substantial differences point to a misfit of the structural part of the model which usually includes more restrictions on the latent levels compared with the CFA (which latent level is saturated by estimating all possible covariances). Hence, some of the restrictions (e.g., zero effects) may be wrong. Hence, rigid evaluation of the model fit (optimally by means of the chisquare test) is vital.
Best,
Holger
Adding to the previous comments, it might be helpful to look at the residual covariances in order to locate the misfit in the model.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling