Causal Inference

Purpose

Estimate causal effects from observational and experimental data using rigorous frameworks and methods: causal frameworks (potential outcomes / Rubin Causal Model, directed acyclic graphs / Pearl's framework, do-calculus, counterfactual reasoning, structural causal models), quasi-experimental methods (difference-in-differences, regression discontinuity design, instrumental variables, propensity score matching, synthetic control), and causal machine learning (uplift modeling, heterogeneous treatment effects, CATE estimation, S/T/X-learners, causal forests, double/debiased ML).

Agent Protocol

Trigger

Exact user phrases: "causal inference", "causal effect", "treatment effect", "potential outcomes", "Rubin causal model", "DAG", "directed acyclic graph", "do-calculus", "counterfactual", "structural causal model", "SCM", "difference-in-differences", "DiD", "regression discontinuity", "RDD", "instrumental variable", "IV", "propensity score", "PSM", "synthetic control", "uplift modeling", "heterogeneous treatment effect", "HTE", "CATE", "conditional average treatment effect", "meta-learner", "S-learner", "T-learner", "X-learner", "causal forest", "double ML", "debiased ML", "confounding", "selection bias", "endogeneity", "identification".

Input Context

Before activating, verify:

• Data source (RCT, observational, panel, time series)
• Treatment assignment mechanism (random, conditional, self-selection)
• Confounders observed and unobserved
• Target estimand (ATE, ATT, CATE, ITE)
• Domain knowledge for DAG construction
• Sample size and dimensionality
• Budget/computational constraints for causal ML