Conflict in results PLS-MGA

Frequently asked questions about PLS path modeling.
Post Reply
lilyv
PLS Junior User
Posts: 9
Joined: Thu Apr 04, 2019 5:07 pm
Real name and title: Lily Vu

Conflict in results PLS-MGA

Post by lilyv » Fri Sep 20, 2019 11:39 am

Hi everyone,
I conduct multigroups analysis with PLS-MGA for 2 groups A & B. The results show that :
Group A : path coefficient = 0.158, p = 0.006 (Sig.)
Group B : path coefficient = 0.063, p = 0.103 (non sig.)
Difference of coeff. = 0.095, p = 0.913.
Objectif is to see if there is significant difference in path coefficient between two groups, if yes, in which group the effect is larger.

According to instruction in Advanced issues in PLS-SEM (Hair et al. 2018), I see that (1-p-value) of the difference is < 0.10 (1-0.913), so the difference beetween 2 groups is significant and the path coefficient is larger in the group B than in the group A. However, for Group B : path coefficient is not significant (It means there is no relationship for group B) and it is lower than group A.

I have some questions about that :
1. Do we compare p-value to 0.10 or 0.05 to assess significant (I think it is 0.05 but why in the book the authors use 0.1) ?
2. Why there is a conflict in these results of path coeff. for group B and p-value of the difference ?
3. Can I conclude that the effect is larger in group B than in group A and there is moderation effect in that case ?

Could anyone advise me how to report these results and tell me where I'm wrong ?

Thank you so much.
Regards,

jmbecker
SmartPLS Developer
Posts: 1045
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Conflict in results PLS-MGA

Post by jmbecker » Thu Oct 10, 2019 9:23 am

2) Are you sure that you compared A against B and not the other way around? In SmartPLS the order of group comparisons depends on how you set them up in the initial dialog and is not alphabetically.

3) Your result would be significant on a one-sided test with 10% probability. If you want to test a two-sided test (more common approach without strict hypotheses about the differences) and an error level of 5% you would need a PLS-MGA p-value above 97.5% or below 2.5% on this test.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

Post Reply