Combining Statistical and Causal Mediation Analysis

Abstract

In this chapter, I argue that mediation analysis practices in social and personality psychology could benefit from the integration of practices from statistical mediation analysis, which is currently commonly implemented in social and personality psychology, and causal mediation analysis, which is not frequently used in psychology. In this chapter, I briefly describe each method on it's own, then provide recommendations for how to integrate practices from each method to simultaneously evaluate statistical inference and causal inference as part of a single analysis. At the end of the chapter, I describe additional areas of recent development in mediation analysis that that social and personality psychologists should also consider adopting to improve the quality of inference in their mediation analysis: latent variables and longitudinal models. Ultimately, this chapter is meant to be a kind introduction to causal inference in the context of mediation with very practical recommendations for how one can implement these practices in their own research.