Dealing with Misspecfication of Formative Variables
Posted: Sat Oct 07, 2017 10:06 am
Hello Collegaues!
I have some doubts concerning this issue as a reviewer has told me that my two formative constructs should emit paths to other unrelated constructs to avoid misspecification in my model. Of course he says it from an AMOS point of view but I do not know how to reply him in PLS, how to demonstrate him that my formative measures have been well defined. i see that a tetrad analysis could offer validity to my formative measures?
Any article or document that can help me to answer him?
I have tested a model with one formative measure related to a reflective (phantom) one, and it works, but i do not know if I must do it for the entire model, I mean, if I have to include in my entire model the phantom variable and also another one unrelated variable for my other formative construct in order to answer him and prove that misspecification is not a problem.
Could I only test individually a model with each formative variable related to a phantom variable and another model with the other formative measure related to an unrelated variable?
Thanks a lot
I have some doubts concerning this issue as a reviewer has told me that my two formative constructs should emit paths to other unrelated constructs to avoid misspecification in my model. Of course he says it from an AMOS point of view but I do not know how to reply him in PLS, how to demonstrate him that my formative measures have been well defined. i see that a tetrad analysis could offer validity to my formative measures?
Any article or document that can help me to answer him?
I have tested a model with one formative measure related to a reflective (phantom) one, and it works, but i do not know if I must do it for the entire model, I mean, if I have to include in my entire model the phantom variable and also another one unrelated variable for my other formative construct in order to answer him and prove that misspecification is not a problem.
Could I only test individually a model with each formative variable related to a phantom variable and another model with the other formative measure related to an unrelated variable?
Thanks a lot