forum.smartpls.com

hi

I'm using smart PLS 3 for my research work , i'm now assessing the outer model , and when I do the cross loading all things are allright horizontally but vertically an item (adoption1) of a latent variable (adoption) loads high on another latent variable (intention) than its first item (intention1) does .
So please would you mind tell me if does this affect discriminant validity ?
or I just have to focus on the horizontal comparison?

I need your advice Dr. Jan-Michael Becker,
thank you in advance

PS:
a sheet of this case is attached below

Usually you interpret the table horizontally. The standard rule is that an item (row) should load higher on its own constructs (column) than on other constructs (columns).
But of course, results like yours can indicate discriminant validity problems. Generally, the cross-loadings of the items of the two constructs are quite high (i.e., around 0.8), indicating problems with discriminant validity. You might also see that the construct have quite high correlations and HTMT and Fornel Larcker might be close to the thresholds.

On the other hand, both constructs also seems to be conceptually very similar if they measure intention and actual adoption. High correlations are expected and desired. Looking at the item wording might also help to decide whether the assignment to the construct makes sense or not.