SMARTPLS 4 Lec 7: Categorical moderator (Binary, dummy variable) Multigroup Analysis and Interaction
The lecture demonstrates two methods for using categorical moderators (binary or dummy variables) in PLS SEM analysis:
Interaction term method - Create an interaction term between the dummy variable and other predictors, then check the significance of the interaction term as well as slope analysis.
Multigroup analysis method - Groups are defined based on the binary variable in the dataset. Then measurement invariance is established through the following steps:
Configural invariance: Establish the same measurement model structure for both groups.
Compositional invariance: Constrain loadings and path coefficients to be equal across groups. If not established, groups should be analyzed separately.
Equality of means and variances: Constrain means and variances to be equal. If established, data can be pooled for multigroup analysis.
For a two-group multigroup analysis, permutation-based testing is recommended for nonparametric analysis.
For groups with large sample size disparity, PLS-MGA (parametric) multigroup analysis can be used, though there is a threat of Type I error.
So in summary, the video walks through two approaches for using categorical moderators in PLS SEM - the interaction term method and multigroup analysis method, outlining the steps for establishing measurement invariance and performing appropriate multigroup tests.