analysis · Claude · free
Hunt for confounders in this observational analysis. Setup: Treatment / exposure: [X] Outcome: [Y] Population + how they were assigned to X vs not X: [DESCRIBE] Covariates observed: [LIST] Domain: [INDUSTRY / SETTING] Deliver: Directed Acyclic Graph (DAG) — as ASCII or bullet edges — with the confounders I should worry about Which observed covariates are true confounders vs mediators vs colliders (adjusting for a collider is worse than not adjusting at all) Unmeasured confounders that would break the claim, with plausibility for each Adjustment strategy (regression, matching, IPW) or why none will save this E value or how strong an unmeasured confounder would have to be to nullify the result What data I should collect next to close the gap
#causal-inference #confounders #dag