10 Causal inference

Chapter 10 provides a comprehensive overview of causal inference methods, beginning with foundational concepts and directed acyclic graphs (DAGs) for identifying confounding structures, followed by core estimation strategies including inverse probability weighting, matching, and the g-formula. It advances to double robust approaches—leveraging SuperLearner, TMLE, and longitudinal TMLE—for machine-learning–enhanced effect estimation, and further covers instrumental variable regression, mediation analysis, confounding and effect measure modification, and the integration of causal frameworks with regression modeling to support valid treatment effect estimation in observational and longitudinal data settings.

10.1 Causal inference introduction

10.1.1 Definitions and directed acyclic graph (DAG)

see here

see here

10.1.1.1 A online tool to draw and analyze causal digrams- DAGitty

see here

10.2 Inverse probability weighting

see here

10.2.1 Matching and weighting w/o imputation

see here

10.3 G formula

see here

10.4 Double Robust Estimators

10.4.1 SuperLearner

see here see here

10.4.2 Double robust estimators using SuperLearner_sl3

see here

10.4.3 Double robust estimators using TMLE

see here

10.4.4 Longitudinal targeted maximum likelihood estimation

see here

10.5 Instrumental variable regression

see here

10.6 Mediation analysis

see here

10.7 Confounding and effect measure modification

see here

10.8 Causal inference and associated regression

see here