Third variables: scientiﬁc meanings and modeling in non-randomized studies
The relationship between two key variables can become considerably more complicated when a third variable is introduced. Third variables can act as moderators, mediators, confounders, and instrumental variables. Their scientific meaning determines their role and how they are modeled. The goal of this chapter is to scientifically distinguish among the various roles of third variables, particularly when attempting to estimate causal effects from data obtained in observational or quasi-experimental studies.