Addressing Causal Heterogeneity In Mediation Analysis: Illusory Indirect Effects As Consequences of Omitted Moderators

Simulations show that omitting moderators can create illusory mediation: indirect effects that are significant for a sample, but not any subgroup. We then identify criteria to help researchers diagnose this problem. Finally, we introduce a novel individual-level index of mediation (using established regression metrics) to recover omitted moderators of mediation.