Mediation test of a formative latent variable

[Christian1988]

Dear All,

I analyze the impact of an IV on a formative latent variable with four reflective indicators. Following Cadogan and Lee (2013), I try to show the relationship between the IV and the formative latent variable at the indicator level of the formative construct and not at the formative latent variable level. In a next step, I analyze the impact of this formative latent variable on a DV.

How can I analyze the mediating impact of the formative latent variable on the relationship between the DV and IV. For a Sobel test of mediation, do I use the beta of the IV for each indicator of the formative construct (as calculated before and then used for several sobel tests) or do I calculate a new beta between the IV and the formative latent variable and then use this for a test of mediation?

I would be very grateful for any help on this!

Thank you very much and kind regards,
Christian

[jmbecker]

You should not use the Sobel test. It is outdated.
You should assess the bootstrapped indirect effect. This should also be readily available for your type of model in the results output.

[Christian1988]

OK, I will do it this way.

Thank you very much for you help, Dr. Becker.

Kind regards,
Christian

[Christian1988]

Dear Dr. Becker,

is there any paper to cite which clearly says that I should not use the outdated Sobel test but should employ the bootstrapped indirect effects?
Thank you very much for your support!

Kind regards,
Christian

[jmbecker]

There are some general paper from Preacher & Hayes discussing this issue for regression analysis. This recommendation has been repeated in the PLS literature, for example, in the second edition of "A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM)" by Hair et al.