forum.smartpls.com

Dear smartpls users,

I am trying to analyse the relationship between two latent variables A--->B.

However, in doing so, I am also trying to control for a possible confounding variable "C" and check for any spurious effects. In plain words I would like to strengthen the hypothesis of causal relationship between variables A and B and make sure that their positive relationship would be there regardless any effect of "C" on both of them.

My question is:
What procedure is best to use in this kind of situation?

Should I compare two different models? The first one with A--->B and the second one introducing the supposed "Confounding" variable "C" pointing at both variables (C--->A, C--->B and A--->B) and then check what happens to the path coefficients A--->B when "C" is controlled for?

Also, C is a single item continuous variable!

Thank you in advance for reading this!
Alex

If you want to show that there is an effect of A on B controlling for C then you should estimate that model including C.
If you want to show that controlling for C does not change the effect of A on B then you need to estimate both models.