One or separate categorical construct/s

[AAljabr]

Hi all,
I have one question.

This question is related to modeling categorical constructs with more than two categories (in my case four). I want to control the influence of this construct. Is it better to include three separate dummy constructs and interpret their results by comparing them with the reference category, or include these three dummy constructs as formative indicators for only formative construct as described in Henseler et al. 2016 paper. However, I have no hint how to interpret the results with later choice. If I use only one formative construct, there will one path coefficient, and I do not know how I can make separate evaluations for the influence of each group.

Can anyone provide advice me on this?

Henseler, J., Hubona, G., & Ray, P. A. (2016). Using PLS path modeling in new technology research: updated guidelines. Industrial Management & Data Systems, 116(1), 2-20.

[jmbecker]

I would add all dummy constructs separately as you would do in a regression analysis. Combining them in a formative construct does not make much sense to me.
What will happen if you combine them in a constructs? Constructs are represented by weighted averages of the indicators in PLS. If you assume that the weights are all equal (why should they be different), then you will have the situation that the construct is 0 if non of the dummies is 1 (ok) and 3 if all 3 are 1 (ok). However, it will be 1 if either one of the dummies is 1 and 2 if two of them are 1. Yet, you will not be able to distinguish which one is asserting its effect at those two cases. Hence, such a construct is masking useful information.

[AAljabr]

Dear Dr. Becker,

Thank you for your clarification.

[Hengkov]

Required new methods and algorithms to handle categorical or ordinal variables in PLS-SEM. Which I mean OPLSc and Non-Metric algortihm.