Page 1 of 1
Size of the groups in PLS Multigroup analysis
Posted: Fri Dec 15, 2017 4:14 pm
by far51
I would like to ask whether there are any conditions regarding the size of the groups for PLS-MGA. I have 70 observations in one group and 200 in the other. There is a huge difference between the two groups in terms of observations and I am not sure if this will affect the reliability of the analysis.
Re: Size of the groups in PLS Multigroup analysis
Posted: Fri Dec 15, 2017 9:39 pm
by jmbecker
We know that it will affect the Permutation approach and therefore also Measurement Invariance testing. The other methods are supposed to be robust against unequal group sizes.
However, 70 observations might generally be too low for a reliable PLS analysis independent of a multi-group analysis.
Re: Size of the groups in PLS Multigroup analysis
Posted: Wed Dec 27, 2017 2:16 pm
by far51
jmbecker wrote:We know that it will affect the Permutation approach and therefore also Measurement Invariance testing. The other methods are supposed to be robust against unequal group sizes.
However, 70 observations might generally be too low for a reliable PLS analysis independent of a multi-group analysis.
Dear Becker, thanks for your answer.
If I am right, SMARTPLS uses permutation approach only?
Would you kindly let me know what other methods can be used for multigroup analysis in PLS?
Moreover, how equal should be the sub-samples sizes for reliable permutation analysis?
Re: Size of the groups in PLS Multigroup analysis
Posted: Thu Dec 28, 2017 10:21 am
by jmbecker
SmartPLS 3 offers you 4 different approaches (2 parametric approaches based on adapted t-tests and a non-parametric approach based on bootstrapping combined in the PLS-Multigroup Feature) and the Permutation approach.
See also
https://www.smartpls.com/documentation/ ... p-analysis
https://www.smartpls.com/documentation/ ... smartpls-3