List the levels of the moderator. For example, if your moderator is the variable “exposure to Western culture,” list “0” and “1,” representing lack of exposure and presence of exposure, respectively.
Count the number of levels of the moderator. In the example, there are two levels.
Copy your structural equation model so that you have a number of models equal to the number of levels in your categorical variable. The model will remain the same at this point.
Add assumptions to your models in line with the different levels of the moderator. For the previous example, you will have two distinct models: one that assumes the subjects have had exposure to Western culture and another that assumes subjects have had no exposure to Western culture.
Divide the subjects into the categories corresponding with the moderator levels. For example, cut the subject data pool into two sets for the previous example: one that contains only subjects with exposure to Western culture and one that contains only subjects with no such exposure.
Run the structural equation models. Run them just as you ran the original model, this time using the data sets that fall in line with the model assumptions.
Compare the resulting models in terms of the relationships between variables. If you find any clear differences between the relationships in the multiple models, there is moderation present. Otherwise, there is no evidence for moderation.