← ClaudeAtlas

analyze-causal-inferencelisted

Estimate causal effects: identification, estimators, robustness bounds.
SylphxAI/skills · ★ 1 · AI & Automation · score 74
Install: claude install-skill SylphxAI/skills
# Analyze Causal Inference Turn a causal claim into an explicit intervention, estimand, identification argument, and falsifiable analysis. Read [references/causal-inference-methods.md](references/causal-inference-methods.md) before choosing a design or adjustment strategy. ## When to use - A claim says one thing causes another and the counterfactual contrast can be stated - You need an identified estimate (not correlation, prediction, or temporal order) with assumptions and sensitivity - Designing or auditing an experiment, holdout, switchback, or observational study of an intervention - Not for competing root-cause hypotheses without an estimand (`analyze-critically`) or feedback dynamics over time (`analyze-system-dynamics`) ## Workflow 1. Define the intervention or exposure, comparator, population, outcome, time zero, follow-up horizon, and target estimand. Reject vague verbs such as Example: "Does the new onboarding flow (intervention) change 7-day retention (outcome) vs the current flow (comparator) for new signups (population) within 30 days (horizon)?" is a complete causal question; "impact" alone is not. “impact” until the counterfactual contrast is clear. 2. Establish temporal ordering and draw the causal assumptions. Distinguish confounders, mediators, colliders, selection mechanisms, measurement error, interference, and time-varying treatment or confounding. 3. Emulate the target experiment conceptually even when only observational data are ava