Primary Submission Category: Causal Discovery
Know Your Role: A Statistical Approach for Distinguishing Mediators, Confounders, and Colliders using Direction Dependence Analysis (DDA)
Authors: Dexin Shi, Amanda Fairchild, Wolfgang Wiedermann,
Presenting Author: Dexin Shi*
In observational data, understanding the causal link when estimating the causal effect of x on y often requires researchers to identify the role of a third variable in the x-y relationship. Mediation, confounding, and colliding are three key third-variable effects that provide different theoretical and methodological implications for drawing causal inferences. However, in practice, these effects are not distinguishable using the commonly used covariance-based statistical methods (e.g., linear regression and structural equation modeling). In this study, we introduce a statistical approach for distinguishing mediators, confounders, and colliders. By using higher-moment information of variables, we propose a two-step procedure within the framework of Direction Dependence Analysis (DDA). Results from Monte Carlo simulations show that our proposed approach accurately recovers the true data-generation process of the third variable. We provide an empirical example to demonstrate the application of our proposed approach in psychological studies. Finally, we discuss the implications and future directions of our work.