Mediation in jasp with categorical predictor
Hi, I am conducting a mediation analysis with JASP. The precitor is categorical (experimental group 1-5). Am I correct to form 4 variables (indicator coding) and put them all into the analysis as predictor variables? How can I interpret the results?
Comments
Hello,
Categorical variables in SEM are tricky, and it really depends on your data and hypotheses you want to test.
In case that the levels in the variable can be ordered, then it is possible to conduct the analysis as if it was a continuous variable, just specifying the type of the variable as "ordinal" should work (in that case the analysis assumes a latent continuous variable).
For strictly categorical data, there are some options, that do not necessarily work for all use cases. At least I am not aware of a standard approach that is suitable for every research question. You could recode the variable as dummy coding indeed, but you need to be careful about interpreting the results (see e.g. https://jslefche.github.io/sem_book/categorical-variables.html for some pointers - note however, that this chapter uses different packages to conduct the analysis than JASP does). In some situations, you could try to rephrase your hypothesis in terms of a multigroup SEM.