DataTerminology.comNEW
Also known as: Causal DAG
A directed acyclic graph encoding causal relationships and conditional independence assumptions among variables.
DESCRIPTION
A causal DAG represents variables as nodes and causal influences as directed edges with no cycles. It encodes assumptions about data-generating processes and supports identification strategies via graphical criteria (e.g., back-door, front-door). DAGs clarify confounding, colliders, and mediation, guiding valid adjustment sets. They underpin do-calculus and are central in modern causal inference. Practitioners use DAGs to design studies, select controls, and avoid bias such as conditioning on colliders.
ALTERNATIVE DEFINITIONS
Wikipedia
A causal diagram is a directed acyclic graph (DAG) that encodes assumptions about the data-generating process.
Focuses on DAGs as diagrams of causal assumptions.
Causal Inference: What If (Hernán & Robins)
DAGs are graphs with directed edges and no cycles used to represent causal relations among variables.
Provides a textbook statement connecting DAG structure to causal interpretation.
RELATED TERMS
Causal Graph Backdoor Criterion Front-Door Criterion Do-Calculus Confounding Collider Instrumental Variable Conditional Independence Structural Equation Model
COMPETING TERMS
KNOWLEDGE AREAS
Data Science AI Mathematics Domain & Business Context
TARGET AUDIENCE
Data Scientist AI Engineer Data Analyst Information Analyst
EXTERNAL SOURCES
- Wikipedia - Causal diagram
- Hernán & Robins - Open textbook on causal DAGs
- UCLA Causal Lab - Pearl’s do-calculus technical report
TOP QUESTIONS
Why use a Directed Acyclic Graph (Causal) in studies?
Causal DAGs reveal appropriate adjustment sets, clarify confounding and colliders, and guide valid identification of causal effects.
Category: Relevancy
How do colliders affect a Causal DAG?
Conditioning on colliders opens spurious paths and introduces bias; DAGs help avoid such conditioning mistakes.
Category: Mistakes
What is a back-door adjustment in a Causal DAG?
Selecting a set of variables that blocks spurious back-door paths from treatment to outcome, enabling unbiased estimation.
Category: Overview
How do DAGs relate to do-calculus?
DAGs provide the graphical structure underpinning do-calculus rules for manipulating interventions and identifying causal effects.
Category: Example
What’s the difference between Causal DAGs and Potential Outcomes?
DAGs are graphical and assumption-oriented; Potential Outcomes formalize counterfactuals. Both frameworks can be complementary.
Category: Alternatives
How are DAGs used in observational studies?
They guide variable selection for adjustment, avoiding bias from conditioning on colliders or descendants of treatment.
Category: Management
What risks occur when DAGs are misspecified?
Incorrect edges or omitted variables lead to invalid adjustment sets and biased causal estimates.
Category: Risks
Can DAGs encode mediation and indirect effects?
Yes; mediator nodes and paths capture indirect effects, enabling front-door identification when assumptions hold.
Category: Overview
How should teams validate Causal DAGs?
Use domain expertise, literature, triangulation, and sensitivity analyses to assess plausibility and robustness of assumptions.
Category: Requirements