DataTerminology.comNEW
A relationship where changes in one variable directly produce changes in another variable or outcome.
DESCRIPTION
Causation denotes cause-and-effect: interventions on one variable change another. Establishing causation requires designs like randomized trials or assumptions modeled with DAGs and counterfactuals. Statistical associations alone do not prove causation. Methods include instrumental variables, propensity score matching, and Pearl’s do-calculus. Causal insights guide decision-making, policy, and product changes, reducing spurious correlations and enabling reliable what-if analysis.
ALTERNATIVE DEFINITIONS
Wikipedia
Causality is influence by which one event or process (a cause) contributes to the production of another event or process (an effect).
Frames causality as influence producing effects, highlighting processes and events.
Encyclopaedia Britannica
Causation, the relation between cause and effect.
Concise statement establishing causation as the cause-effect relation.
RELATED TERMS
Causal Inference Randomized Controlled Trial Directed Acyclic Graph Counterfactual Do-Calculus Instrumental Variable Propensity Score Matching Uplift Modeling Directed Acyclic Graph (Causal)
COMPETING TERMS
KNOWLEDGE AREAS
Data Science Mathematics AI Governance
TARGET AUDIENCE
Data Scientist Data Analyst AI Product Manager Data Strategy Consultant Chief Data Officer
EXTERNAL SOURCES
- Wikipedia - Causality
- Stanford Encyclopedia of Philosophy - Philosophical overview of causation
- Judea Pearl - Foundational causal inference resources
TOP QUESTIONS
What is causation in data science?
Causation means changes in one variable produce changes in another. It underpins policy evaluation, product changes, and reliable decision-making.
Category: Definition
How does causation differ from correlation?
Correlation is association; causation is cause-and-effect. Associations can be spurious, while causal links require design and assumptions.
Category: Misunderstanding
Which methods establish causation?
Randomized controlled trials, instrumental variables, propensity score matching, and causal DAG modeling with do-calculus are common approaches.
Category: Overview
What is currently a hot topic around causation?
Causal machine learning, uplift modeling, and integrating causal graphs with deep learning are active research areas.
Category: Topic
What are common mistakes when inferring causation?
Treating correlation as causation, ignoring confounders, and weak instruments cause biased estimates and wrong conclusions.
Category: Mistakes
How is causation applied in product decisions?
A/B tests estimate causal uplift. Teams use experiment results and causal models to decide feature rollouts and marketing changes.
Category: Example
What are risks of misidentifying causation?
Misidentified causes lead to wasted investments, ethical harms, and poor policy. Carefully validate assumptions and perform sensitivity analyses.
Category: Risks
When was modern causal inference popularized?
It matured in the late 20th century, with Judea Pearl’s graphical models and do-calculus widely influential in the 1990s and 2000s.
Category: Origin
Where should I start learning causation?
Begin with Pearl’s works and practical guides, then explore DAGs, RCTs, and applied tutorials using causal libraries.
Category: Requirements