I would like to know when choosing a mediator variable, researchers should consider its theoretical relevance to the IV and DV? Thus, is it always the case that, the selected mediator should logically account for the relationship between the independent and dependent variables and demonstrate a significant association with both the independent variable and the outcome?