Discipline II · INSECAB Analytical Framework
Correlation describes associations. Causality makes it possible to reason about interventions. In medicine and public health, confusing the two can lead to mistaken conclusions with real consequences.
Biomedical analyses frequently present associations without explicitly distinguishing them from causal relationships. Causal inference provides the formal framework for making that distinction and for stating precisely what a study is claiming.
At INSECAB, causal inference runs across the whole of the training. The aim is that every analytical decision — from building a DAG to interpreting the results — has an explicit causal justification where appropriate.
The formal framework of causality
Potential outcomes (Rubin)
A counterfactual framework for defining the causal effect of an exposure rigorously.
Do-calculus (Pearl)
A causal algebra for reasoning about interventions in complex systems.
Double robustness
Estimators designed to deliver consistency if at least one of the two auxiliary models is correctly specified.
Bias vs confounding
Precise diagnosis of threats to internal validity before any analysis.
Methods and applications
Explicit causal reasoning before modelling. Identifying confounders, mediators and colliders in order to choose which variables to adjust for and which to leave alone.
Confounding control in observational studies. Estimation of causal effects through matching, IPTW weighting and propensity score stratification.
A doubly robust estimator for the average treatment effect (ATE) and the effect on the treated (ATT). Combining machine learning with semiparametric influence function theory.
Regression discontinuity, difference-in-differences and instrumental variables for causal inference when randomisation is not possible.
Decomposition of total effects into direct and indirect effects. Identifying the causal mechanisms that explain an observed association.
Other disciplines in the INSECAB framework