Validity and reliability of formative construct

[klanias]

Hello everyone,
so I have developed a model in which all but two variables are reflective. For the reflective constructs I am aware of the tests of reliability and validity as well on how to report results. My question is how to ensure validity and reliability of formative constructs, and how to report these.

My predictor construct is a formative construct with 12 items in which weights are not significant towards the variable itself. I have seen that this is not a problem, and unlike reflective constructs I shouldn't proceed in deleting items that are non-significant.

I also have a second order construct (formative-formative), how do I also determine validity and reliability of this variable and its sub-variables?

Finally, how do I present results in a uniform manner. I know that when all constructs are reflective a matrix is developed with sqrt of AVE in the diagonal and construct correlations below the diagonal. Also CR is used. What do I do in the case of formative measures for which CR and AVE are 0?

Thanks in advance :)

[Hengkov]

Hi,

CR and AVE not appropriate for mode B. Used Nomological validity

Regards,

[bradley7680]

Agreed. CR, Alpha and AVE are a nonsense for formative constructs.

The overall quality and predictive ability of the model is best test.

brad

[gulmmerve]

Agreed. CR, Alpha and AVE are a nonsense for formative constructs.

The overall quality and predictive ability of the model is best test.

brad

Mr. Brad, I was searching for "Is measuring AVE valid for formative constructs" and came across with your post.

I would appreciate if you could share a reference paper for your above comment.


Regards,

Merve

[cringle]

No, it's not.

Best
CR