Page 1 of 1

Multicollinerity for 2nd order formative construct

Posted: Fri Jan 26, 2018 12:43 pm
by shash
Hi,

I face multicollinearity issue with my 2nd order formative construct. after reading some forum tracks, i used the 2 step approach. I used the original IVs with the original items and used the Latent Variables (results in reports) generated via running the PLS Algorithm for the DV (2nd order construct). With this model, the VIF readings are ok.

I just wanted to confirm whether the way i did this correct? Using the original construct and items for IVs and other reflective constructs except for the DV which is 2nd order formative, using the Latent Variables generated from SmartPLS. I used all the Latent Variables generated for the 1st order construct of the DV (means not a single item).

Please help. Tq. :)

Re: Multicollinerity for 2nd order formative construct

Posted: Mon Jan 29, 2018 1:48 pm
by jmbecker
With the given information it is hard to assess, whether it is correct or not. Please provide a more detailed explanation of your approach.
In particlar, how did the model look like from which you took the LV scores to be items for your second-order construct.
Where did you have multicollinearity problems that were solved after using the two step approach?