Chapter 7 integrates direction of dependence properties of variable distributions, error distributions, and predictor–error independence patterns into a cohesive framework. The resulting framework involves six distinct steps, and enables researchers to empirically test hypotheses of causation and derive statements about the direction of the underlying causal mechanism that generates the observed variable association. Furthermore, the chapter discusses principles of covariate selection in modeling the direction of dependence. These principles can serve as the basis for identifying eligible and rule out non-eligible covariates, that is, empirically distinguishing between confounders (common causes which should be adjusted for) and colliders (common effects which should not be adjusted for to avoid biased conclusions). In addition, we provide an exposition of the assumptions necessary to perform Direction Dependence Analysis (DDA) and discuss potential remedies in cases of assumption violations. The chapter concludes with an illustration of the steps taken in DDA using a real-world data example.