As a novice with Mplus, I have a question concerning the identification of a second order factor model.
I have a second order factor model consisting of only 2 first order factors (f1 and f2), that influence the second order factor (formative factors; f1 -> f etc.). The first order factors are measured by 4 variables (x1-x4, X5-x8) each (reflective indicators; f1 -> x1 etc.).
From a theoretical perspective my model makes sense, but, obviously, I have a serious identification problem. Is there a reasonable way to achieve identification? Does it help in any way that I also have a one-item-measure for f (second order factor)?
Dodam Park posted on Wednesday, April 05, 2017 - 3:30 am
I have the same problem like Stefanie Exler did. In my model, the first order consist of the two second order factors. If I can't identify this model, then would it be okay to just combine all the factors indicating the one latent construct instead of distinguishing the first order factor and the second order factors? I wonder if this would be an acceptable idea as long as I have substansive theoritcal reasons.