Hello, I am using SmartPLS 3.2 and want to analyse especifically the change of interaction terms (predictor-moderator) with multi-group analysis. Unfortunately, I still have some serious doubts - hopefully you can help me:
1) when splitting my sample into groups of high (n=51) and low (n=49) a small section of "middle" rests (n=10) -->should this middle section be eliminated for the analysis BEFORE the mga in order to make conclusion with and without the mga?
2) is it possible to easily check for measurement invariance with the smartpls 3.2.0 version?
3) unfortunately, PLS-MGA does not check signifance of the differences between the two groups concerning the interaction terms (which is I urgently want to test): the Welch-Satterthwaite test is available at the MGA section but just refers to parametric tests. So would you recommend this one (provided that there is measurement invariance- question 2))?
OR: is it possible to calculate by hand the significance of the interaction terms using the path coefficents and Confidence Intervals available in the result section?
Thank you so much
MGA - Interpretation of interaction terms
-
cringle
- SmartPLS Developer
- Posts: 818
- Joined: Tue Sep 20, 2005 9:13 am
- Real name and title: Prof. Dr. Christian M. Ringle
- Location: Hamburg (Germany)
- Contact:
Re: MGA - Interpretation of interaction terms
Hi
@1: This has nothing to do with SmartPLS but how you design the MGA. I usually leave the middle third our if I have enough data.
@2: How do you want to test measrument invariance in PLS?
@3: PLS-MGA gives you the significance of group differences. Are you running a moderator anaylis (intercation term), which is a form of MGA, and a MGA at the same time? Sounds strage. If you really want to do this, you need to run bootstapping for each group and hand calculate the results.
Best
CR
@1: This has nothing to do with SmartPLS but how you design the MGA. I usually leave the middle third our if I have enough data.
@2: How do you want to test measrument invariance in PLS?
@3: PLS-MGA gives you the significance of group differences. Are you running a moderator anaylis (intercation term), which is a form of MGA, and a MGA at the same time? Sounds strage. If you really want to do this, you need to run bootstapping for each group and hand calculate the results.
Best
CR
Prof. Dr. Christian M. Ringle, Hamburg University of Technology (TUHH), SmartPLS
- Literature on PLS-SEM: https://www.smartpls.com/documentation
- Google Scholar: https://scholar.google.de/citations?use ... AAAJ&hl=de
- Literature on PLS-SEM: https://www.smartpls.com/documentation
- Google Scholar: https://scholar.google.de/citations?use ... AAAJ&hl=de
-
wittmannv
- PLS Junior User
- Posts: 7
- Joined: Tue Feb 24, 2015 1:37 pm
- Real name and title: Verena Wittmann
Re: MGA - Interpretation of interaction terms
Hi, thanks a lot for your reply!
@1: Yes, sorry, I know, I have misplaced the question. It should have belonged to the methodology part of the forum or even a PLS independent section. But thanks anway. One last question to this: the elimination of the middle section from the whole sample (i.e. no analyse with this part in the MGA indepedent section) is justifyable for the reason that a comparison of the models with and without group analysis is possible, whereas the big disadvantage is a loss in power?
@2: No, the question was intended to consult you if I have missed out any possibility of doing a messinvariance test because I neither know any possibility of checking this.
@3: For calculating it manually, I have used the formula which is available in the Hair et al. (2014, p. 248) book for the parametric test, and the formula in Sarstedt et al. (2011). The one for parametric tests works out fine but I have difficulties with the one for the PLS-MGA: as it is said, one one-sided tests can be analysed. Anything that has to be kept in mind?
You said you would calculate the two groups in different submodels, isn´t it possible within one model with the formula?
Thank you very much!
@1: Yes, sorry, I know, I have misplaced the question. It should have belonged to the methodology part of the forum or even a PLS independent section. But thanks anway. One last question to this: the elimination of the middle section from the whole sample (i.e. no analyse with this part in the MGA indepedent section) is justifyable for the reason that a comparison of the models with and without group analysis is possible, whereas the big disadvantage is a loss in power?
@2: No, the question was intended to consult you if I have missed out any possibility of doing a messinvariance test because I neither know any possibility of checking this.
@3: For calculating it manually, I have used the formula which is available in the Hair et al. (2014, p. 248) book for the parametric test, and the formula in Sarstedt et al. (2011). The one for parametric tests works out fine but I have difficulties with the one for the PLS-MGA: as it is said, one one-sided tests can be analysed. Anything that has to be kept in mind?
You said you would calculate the two groups in different submodels, isn´t it possible within one model with the formula?
Thank you very much!