Hi all,
I know how to do multi group analysis in PLS-SEM for two groups. Regarding more than two groups I read a number of papers and I reached this dummy coding. the method is straight forward when it is talking about the dummy variable and I think I have no problem with that.
For instance this link is how it works (http://www.ats.ucla.edu/stat/mult_pkg/faq/general/dummy.htm). But at the end, I cannot understand how they use this method to analyse moderators.
Is it true if in a simple language I say: "we take one group as the reference and compare others one by one with that particular groups"? if it is true, why we bother ourselves to use dummy variable?
Thanks in advance,
Instead to apply the dummy variable for multi-group analysis more than 2 groups, i prefer you to apply permutation test. This method is one of the non-parametric test that has a capability to handle more than 2 groups. You can find the step of this approach by Chin paper.
I was curious how does the dummy variable works.
Thanks anyway. I will read about what you propose
Hosein, you would not use both methods simultaneously in your case, if I understand you correctly. First, you could perform a multi-sample analysis and constrain one parameter or a few parameters of interest to be invariant across groups (if this is possible in PLS). Second, you could just specify a single SEM (for all subjects of all groups) and include dummy variables for group membership or the difference between a group of interest and a comparison group, respectively.
Using multi-sample analysis would be a good idea in order to test for parameter invariance across groups.
HTH, Karin
Current thought in the statistical literature is that PLS rarely offers any benefits over covariance-based SEM and/or OLS regression modeling, see for example Rönkkö et al. (2014) – see http://dx.doi.org/10.1016/j.jom.2016.05.002 – who judge PLS very harshly. So, try not to use PLS at all.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series