Triple 2nd Order

[Jako]

So this is my Initial Model with three 2nd Order Constructs:



All Items that are in the 1st order constructs are additionally in the 2nd order (how I think it is supposed to be).

I started deleting items that do not fit (correlation <0,7) from 1st and 2nd order, then created an Excel with the results from the Report "latent variables", deleted 1st order constructs and inserted the rest of the data into my data csv.



First question: Is it correct to Keep the one-dimensional constructs? Or do I have to delete them?

Then, i created a new Project with the new file.



Have I done everything correct? Do I Need to put in the Items for the one-dimensional constructs or the now newly created "summarizing item"?

Sorry, I am rather new using SmartPLS. Thanks for advice.

[cringle]

We are not big fans of higher-order constructs (they often times make no sense content-wise). You use the repeated indicator approach. Reusing all indicators of the lower order constructs in the higher order construct makes sense. But do not delete low loading item. Rather relation from the higher order construct to the lower order constructs represents the loadings.

Kind regards
Christian Ringle

[Jako]

Thank you for the answer, Prof. Dr. Ringle.

As I've seen in one of your books, A Primer on PLS-SEM, I have two types here, reflective-formative and reflective-reflective. As my theory work suggests, it makes sense to keep those created 2nd order constructs and not bringt it down to one Level.

Besides Formalization, the model perfectly works when I put all items on the 2nd order and delete first order (see Picture 1 & 2). My question now is, how can I detect the same validity and reliability indicators when using the 2 stage approach.

And: Should I delete the 1st order latent variables in the 2nd step of the 2step Approach (i.e. DEA, REA ...)?

Again, I am very thankful for your advice.

Best regards
Kohlen

Picture 1:


Picture 2:


Picture 3:

[cringle]

To better understand how to conduct the Two-Stage approach, take look at this article (Appendix):
Ringle, C.M., Sarstedt, M., Straub, D.W., 2012. A Critical Look at the Use of PLS-SEM in MIS Quarterly. MIS Quarterly 36, iii-xiv.
http://misq.org/skin/frontend/default/m ... sV36N1.pdf

You use the LOCs latent variable scores as indicators of the HOC in the second stage (and yes, you delete the LOCs since the are used as manifest variables in the measurement model of the HOC). You get the evaluation criteria of the HOC directly from the report.

Best regards
CR

[Jako]

Again, thank you so much for your time and effort.

Best
Kohlen