I'm doing a multi-group analysis with MPlus. In addition to other variables, the model contains two latent constructs. With respect to these two constructs, I was wondering if the factors are rotated or not.
If rotation occurs, can the type of rotation be influenced? If not, could this be a possibly problematic limitation of MPlus?
In CFA, the default is that all factor correlations are freely estimated in Mplus, that is, the factors are "oblique." There is no factor rotation issue in CFA. You can change the Mplus default to fix the factor correlations a priori to whatever your theory predicts (e.g. to zero for theoretically uncorrelated factors).
When performing EFA in Mplus, you can pick the rotation method like in other programs for EFA.
To add one more possibly relevant idea to the above (correct) answer: Although it is unlikely that the freely estimated correlation between factors will be precisely zero, it could be very small and non-significant. In which case you might wish to test whether an orthogonal solution is an equally good fit to your data. This can be done by introducing the constraint on the size of the correlation between the factors to be exactly zero. Since they are nested models, there is a direct test of the difference in level of fit between the two models.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling