Hi everyone,
I've come across several articles that raise concern about CFA and SEM fit indices pointing out that a mis-specified model can have excellently fitting indices. If you have a 6 item scale and accept the recommendation that one factor should be determined by at least three variables, there seem to be only two model options a 1 and a 2 factor solution (3-3 variables). Can this be mis-specified in any way? Sorry, if I'm asking something trivial, I'm not good at stats. Thanks for any comments in advance.
Kind regards
Attila
In order for it to accept these two models that it puts of example it has to be of agreement to a theoretical model, where the factorial vargas must be greater to .50, also it must have in consideration the distribution of multivariate normality of Mardia, in case it does not fulfill it must make the correction of Satorra-Bentler, also in a possible bad adjustment it can have to see that a same source of error between indicators (similar writing, conceptual, or correlation) reason why the covariance of errors between these indicators can be made.
- Badges
- Science topic
- Similar topics
- Comment