Multigroup analyses

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
Choukri
PLS Expert User
Posts: 114
Joined: Thu Apr 16, 2015 2:11 pm
Real name and title: Choukri MENIDJEL

Multigroup analyses

Post by Choukri »

Hi,

To make a multigroup analyses by gender, I divided my research sample to two data sets using SPSS (Female and male), than I used SmartPLS 3 to test my model depending on the two data sets, after runing SmartPLS, I got two models with defferent results.

Is this way correct to make multigroup analyses by gendre or age?
What is the statistics should I get from SmartPLS to justify the defference between the two models?

SmartPLS Team, thank you
User avatar
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: Multigroup analyses

Post by cringle »

First, you do not need to use SPSS to split your data set. SmartPLS has in-built options to create groups of data. Double click on your data set and use the group generation functions in the menu.

Second, when runnding the multigroup analyis, you get the results of several tests that show you if the difference is signifficant. E.g., when using PLS-MGA: All probabilities above 0.95 and blow 0.05 indicate that the group differences are significant.

Regards
Choukri
PLS Expert User
Posts: 114
Joined: Thu Apr 16, 2015 2:11 pm
Real name and title: Choukri MENIDJEL

Re: Multigroup analyses

Post by Choukri »

Thank you Pr. Christian M. Ringle

You are so helpful

Thank you
Choukri
PLS Expert User
Posts: 114
Joined: Thu Apr 16, 2015 2:11 pm
Real name and title: Choukri MENIDJEL

Re: Multigroup analyses

Post by Choukri »

Hi,
Take a look on this

viewtopic.php?f=12&t=3405&p=12173#p12173

Best regards
ckoch1979
PLS Junior User
Posts: 1
Joined: Mon Mar 14, 2016 3:13 pm
Real name and title: Christian Koch

Re: Multigroup analyses

Post by ckoch1979 »

Hallo Herr Prof. Ringle,

muss der Wert definitiv unter 0,05 und über 0,95 oder kann er auch gleich 0,05 oder 0,95 sein?

Kann das Signifikanzlevel eigentlich im Bootstrapping-Dialog für die MGA geändert werden? Wenn ich es auf 0,02 ändere, sind dann bspw. aller Werte unter 0,02 und alle Werte über 0,98 signifikant?

Viele Grüße
Christian Koch
MrcJnk
PLS User
Posts: 16
Joined: Thu Oct 17, 2013 10:04 am
Real name and title: Marc Janka
Location: Germany

Re: Multigroup analyses

Post by MrcJnk »

Dear Mr. Koch:

The level of significance should be set on the basis of your target journal. 5% is more conservative but there are also journals that accept higher levels up to 10%. The setting in SmartPLS is primarily for calculating the confidence intervalls. You can see the change in the output report in the head of confidence intervalls tables.

Best regards

MJ
PATR
PLS Junior User
Posts: 3
Joined: Tue Aug 02, 2016 1:49 pm
Real name and title: Patrick R

Re: Multigroup analyses

Post by PATR »

Hello community! : )
after checking my research model with the datasets of all participants (81) of my survey, now I want to further evaluate whether there are differences between male and female participants.

Here, first I created the new groups "female"(22) and "male" (59). Then I started the PLS Algorithm, applied the user groups and was able to switch between ALL/female/male. As I need the T-values, I also started the Bootstrap calculation and also here I applied the user groups but now, I just can switch between ALL/male.
Do somebody know why female is missing and how do I get the results for that?

I also tried to start the Multi-Group Analysis but I get the error message "An error occured during the outside estimation for latent variable "BIU". (Double click for more information!)". Here the double click says remove the variable or increase sample size - both is no opportunity.
Any suggestions how to solve the problem?

thanks and best,
Patrick
Brusset_X
PLS Junior User
Posts: 9
Joined: Tue Feb 19, 2013 4:05 pm
Real name and title:

Re: Multigroup analyses

Post by Brusset_X »

I also tried the Multi-Group analysis and get the same error message, and when I click on it, I get the following description:
"There may be too few observation to consistently estimate the regression coefficients in a Mode B measurement model or extreme collinearity "

It would appear that groups (or subgroups) of less than 20 cases generate this error. I wanted to test whether the answers given by respondents where different according to their hierarchical position (1 stands for head of business unit, and 5 for operator at least 4 levels under head). As I have 2 sub-groups with only 16 and 20 cases, I had to take them both out for the analysis to run to the end. To be complete, I did 2 groups: both composed of two subgroups, but excluding the offending ones.

Can somebody explain why this is so and present some explanation as to how one can do?
Thank you.
Xavier Brusset
Professor
Logistics and supply chain management
Toulouse Business School
jmbecker
SmartPLS Developer
Posts: 1282
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Multigroup analyses

Post by jmbecker »

Your subgroups are simply too small. It is often not possible to estimate complex PLS models with only 16, 20, or 22 observations. And even if it would be possible the results would not be trustworthy. In your cases a multi-groups analysis is simply not possible unless you collect more data.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Post Reply