Getting confused about the measurement model for MGA and higher-order factor

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
PLS Junior User
Posts: 1
Joined: Wed Dec 21, 2022 8:01 am
Real name and title: Theerapong B.

Getting confused about the measurement model for MGA and higher-order factor

Post by al4872 »

Dear PLS-SEM experts,

I am new to this analytical approach and have some questions as follows:
I have one complete group (whole group) and this group is later divided into two groups that will be used for the multi-group analysis. My question is, when reporting the measurement model in a research article, do I have to present the reliability and validity (e.g., loadings, AVE, CR, alpha, HTMT) for both groups (that will be used for MGA) or just for the whole group sample. I found that all the reliability and validity values for the whole group qualified without problems, but they were below the threshold when for the two separate groups (especially factor loadings). Is this normal situation?
*n for the first group is about 120 and about 170 for the second group.

My second question is that my model contains three higher-order factors. The values such as CR, AVE, Cronbach’s alpha are always low for the higher-order factors regardless of the complete group sample or the splitter group samples. I am curious if, in SmartPLS4, I still have to calculate these values using the formula given in the following article “Sarstedt et al. (2019). How to specify, estimate, and validate higher-order constructs in PLS-SEM”? (My model is reflective-reflective)

Thank you in advance for all answers and suggestions

Post Reply