Hi,
Before running my full size path model, I tested the convergent and discriminant validities separately.
For convergent validity, I developed LV1 --> LV1 models for all LVs individually, and everything was fine (high loadings, etc.)
For discriminant validity, I developed LV1--> LV2 models for all LVs separately to compare the AVEs with the square of the Correlation between the LVs and also checking the cross loadings.
The problem is that, for some LVs, the excellent loadings I had in LV1-->LV1 models does repeat in the LV1 --> LV2 models, and few of them are extremely poor (lass than 0.5), and I have no idea why this happens !!
Changing Loadings
Changing Loadings
Soroosh Saghiri