PLS-MGA with dependent samples

PLS is broadly applied in modern business research. This forum is the right place for discussions on the use of PLS in the fields of Marketing, Strategic Management, Information Technology etc.
Post Reply
PLS Junior User
Posts: 7
Joined: Tue Jun 09, 2009 2:40 pm
Real name and title:

PLS-MGA with dependent samples

Post by Taolin » Tue Apr 18, 2017 9:01 am

Hi everyone

I got a question, related to PLS-MGA and dependent samples.

Since PLS-MGA seems to be designed for comparing different groups (e.g., genders, cultures), I assume that the results are computed based on a independent sample t-test.

I collected panel data, and my goal is to investigate whether the strength of the path coefficients changes over time (e.g., some antecedents becoming more/less important over time). Now, the thing is that my samples are not independent, since the exact same participants completed the survey multiple times (3 weeks in between).

My question is whether I could still run the PLS-MGA, and then take parts of the output (e.g., interim results) to compute a dependent sample t-test for my purpose. If yes,a) which parts of the output do I need, b) which formula should I use with this output, and c) can I just use the basic PLS-MGA-settings or do I need to make changes?

Thanks in advance,

SmartPLS Developer
Posts: 821
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: PLS-MGA with dependent samples

Post by jmbecker » Sun Apr 23, 2017 12:34 pm

Simple answer: There is no established method to account for dependent samples in MGAs, although there are probably many possible ways to account for this. Certainly a research opportunity.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: ... v=hdr_xprf
GoogleScholar: ... AAAJ&hl=de

PLS Junior User
Posts: 1
Joined: Fri Dec 29, 2017 2:46 pm
Real name and title: Jeff Mullins

Re: PLS-MGA with dependent samples

Post by jmullins » Fri Dec 29, 2017 3:09 pm

Matthias, thanks for posting this question. Did you ever determine a good solution to the problem you described here with dependent samples and PLS-MGA? I am currently facing the same issue, wanting to compare path coefficients in a PLS model before and after an intervention. And thank you, Dr. Becker, for clarifying what Matthias suspected and confirming that MGA is not set up to account for this by default.


Post Reply