Discriminant Validity Question

[paolas25mx]

Hi everybody,

I have a model with 2 reflective latent variables and 7 formative, when I evaluate the discriminant validity of the reflective models, I am comparing:

1) I compare the square root of AVE against the correlation of the reflective latent variable with all the other reflective and formative constructs.
But I have read a paper on the Handbook of PLS, where they exclude the correlations with the formative constructs, can I include it? Cause it happens that some correlations with formative constructs and the reflective one are higher than the square root of AVE (of the reflective one)

2) I compare the crossloadings, but again, I left the crossloadings of the reflective indicators in the formative ones; it happens that the reflective indicators of one construct have very high loadings on the formative constructs. Can I leave like that, or should I compare exclusively with reflective constructs?

Thanks so much

Ana Paola Sánchez Lezama