Causal Inference From Observational Data Chain

validated scientific chain v1.0.0 cc-by-sa

Block ID: 26f72013-7c02-4ee0-8d9e-9dee576e408e

Community-contributed block. PromptDNA makes no guarantee of output quality or fitness for purpose. User assumes all responsibility for use.

Template

Evaluate whether the observed association in {association_description} plausibly reflects causation. Work the sequence: (1) State the association precisely — direction, magnitude, and the population it was observed in. (2) Enumerate the rival explanations before advocating for causation: chance (what's the statistical strength?), reverse causation (could the outcome drive the exposure?), confounding (list specific plausible third variables, not just the abstract possibility), and selection or measurement bias in how the data arose. (3) For each named confounder, assess whether the analysis adjusted for it, and whether adjustment was even possible if it was unmeasured. (4) Apply the classic considerations where relevant — strength, dose-response, temporality, consistency across independent studies, mechanistic plausibility — treating them as evidence to weigh, not a checklist that certifies causation. (5) Identify what evidence would discriminate: a natural experiment, an instrumental variable, or an intervention study. (6) Conclude with a calibrated verdict — from "association only" to "causation well supported" — and the single alternative explanation that remains least excluded.

Evaluate whether the observed association in {association_description} plausibly reflects causation. Work the sequence: (1) State the association precisely — direction, magnitude, and the population it was observed in. (2) Enumerate the rival explanations before advocating for causation: chance (what's the statistical strength?), reverse causation (could the outcome drive the exposure?), confounding (list specific plausible third variables, not just the abstract possibility), and selection or measurement bias in how the data arose. (3) For each named confounder, assess whether the analysis adjusted for it, and whether adjustment was even possible if it was unmeasured. (4) Apply the classic considerations where relevant — strength, dose-response, temporality, consistency across independent studies, mechanistic plausibility — treating them as evidence to weigh, not a checklist that certifies causation. (5) Identify what evidence would discriminate: a natural experiment, an instrumental variable, or an intervention study. (6) Conclude with a calibrated verdict — from "association only" to "causation well supported" — and the single alternative explanation that remains least excluded.

Variables

Name	Type	Required	Trust level
association_description		yes

causal-inferenceconfoundingobservational-studieschain-of-thought

Ratings

0.0

Overall (0)

0.0

Accuracy

0.0

Consistency

0.0

Clarity

0.0

Efficiency

Benchmarks

Not yet self-validated against any benchmark. Automated, evaluative only - not a factor in whether this block was published.

Submitted by James P FounderMod via mcp · 2026-07-18