UM Data Science Research Seminar with CAPHRI

Schedule

LECTURE (12:00 -12:45) - Q&A (12:45 - 13:00)

Speaker: Wen Wei Loh

Subject: "Causal mediation analysis for multiple mediators with interventional indirect effects" 

Abstract
Mediation analysis is a popular tool in health, medical, and social sciences. However, routine approaches have been heavily criticized. How can we fortify causal inferences from mediation analysis in practice? In this talk, I introduce a novel and elegant approach from the causal inference literature: interventional indirect effects. Using Pearl’s graphical models (i.e., Bayesian networks) framework, I will explain why standard methods such as serial mediation models rely on strict and often overlooked assumptions that can lead to incorrect conclusions. In contrast, interventional indirect effects avoid these pitfalls by focusing on mediator-specific causal effects. No assumptions about the mediators’ causal or confounding structure are needed. Examples of recent applications in health and social sciences are listed below.

1. American Journal of Epidemiology: Causal effects of competing obstetrical interventions: mediators of placental abruption and perinatal mortality (doi.org/10.1093/aje/kwae273)
2. American Psychological Association (APA) Psychnet, issue 'Emotion': Do cues of infectious disease shape people’s affective responses to social exclusion? (doi.org/10.1037/emo0001157)
3. Epidemiology: Does Adjusting for Causal Intermediate Confounders Resolve the Perinatal Crossover Paradox? (DOI: 10.1097/EDE.0000000000001848)
4. Journal of Affective Disorders: Explaining how a psychosocial intervention (PROACTIVE) based on behavioural activation improved outcomes of depression in older adults living in deprived regions of Brazil: The mediating roles of reduced loneliness and stepped care. (doi.org/10.1016/j.jad.2024.12.024)
5. Pain: Reductions in kinesiophobia and distress after pain neuroscience education and exercise lead to favourable outcomes: a secondary mediation analysis of a randomized controlled trial in primary care. (DOI: 10.1097/j.pain.0000000000002929)