[Three-level meta-analysis] how to integrate all sort of variables in a model?