DataTerminology.comNEW
Also known as: SCM
A formal causal framework where variables are generated by structural equations and exogenous noise under a directed acyclic graph.
DESCRIPTION
A Structural Causal Model (SCM) specifies how each variable is determined by its parents and an exogenous noise term via structural equations. Together with a DAG, SCMs define observational and interventional distributions, enabling causal queries, counterfactual reasoning, and identifiability via do-calculus. SCMs unify causal graphs with mathematical semantics, supporting both estimation and reasoning tasks.
ALTERNATIVE DEFINITIONS
Wikipedia
A structural causal model is a collection of structural equations and a causal graph that together define the causal semantics of a system.
Emphasizes equations and graph jointly defining causal semantics.
Judea Pearl (UCLA)
An SCM consists of endogenous variables determined by functions of their parents and exogenous noise variables under a causal graph.
Provides the structural-equation-based definition with endogenous and exogenous variables.
RELATED TERMS
Causal Graph Directed Acyclic Graph Do-Calculus Treatment Effect Counterfactual Structural Equation Backdoor Criterion Front-Door Criterion Identification g-Formula
COMPETING TERMS
RELATIONS
- Broader: Model
KNOWLEDGE AREAS
Data Science Mathematics AI Semantic & Knowledge Modeling
TARGET AUDIENCE
Data Scientist AI Engineer Information Analyst Data Strategy Consultant
EXTERNAL SOURCES
- Wikipedia - Structural causal model
- Brady Neal - Intro text covering SCM basics
- Elements of Causal Inference (Peters et al.) - Open textbook on causal models
- UCLA Causal Research - Pearl’s causal modeling materials
TOP QUESTIONS
What is a Structural Causal Model (SCM)?
An SCM defines variables via structural equations and exogenous noise under a DAG, enabling causal and counterfactual reasoning.
Category: Definition
How do SCMs relate to DAGs?
DAGs express the causal structure; SCM equations define how variables are generated consistent with that structure.
Category: Example
What is the role of exogenous variables in SCMs?
They capture noise and unmodeled factors, assumed mutually independent under standard formulations.
Category: Definition
How do SCMs support identifiability?
They enable do-calculus and graphical criteria (e.g., backdoor) to derive estimands from observational data.
Category: Application
What risks occur when using SCMs?
Misspecified graphs or functional forms can yield biased conclusions; assumptions must be justified and tested via sensitivity analysis.
Category: Risks
How do SCMs compare to the Potential Outcomes Framework?
Both target causal effects; SCMs use explicit structural equations and graphs, while potential outcomes focus on design and ignorability assumptions.
Category: Alternatives
Can SCMs answer counterfactual questions?
Yes. SCMs naturally extend to counterfactual queries via abduction–action–prediction steps.
Category: Example
Where to learn SCMs?
Pearl’s Causality, Neal’s introduction, and Elements of Causal Inference provide accessible yet rigorous coverage.
Category: Learning
Are SCMs only for academics?
No. They inform practical decisions in tech, healthcare, and policy by clarifying assumptions and guiding valid adjustments.
Category: Relevancy