DataTerminology.comNEW
A set of algebraic rules for transforming causal expressions to identify interventional effects from observational data given a causal graph.
DESCRIPTION
Do-calculus, introduced by Judea Pearl, comprises three inference rules for manipulating expressions involving do(·) operators in structural causal models. It determines whether interventional effects are identifiable from observational distributions given a DAG. The rules enable deriving adjustment sets and front-door/backdoor solutions. It underpins algorithmic approaches to causal identification and automation in tooling.
ALTERNATIVE DEFINITIONS
Wikipedia
Do-calculus is a set of rules of inference that allow one to transform probabilistic expressions with interventions into purely observational expressions.
Emphasizes transformation from interventional to observational distributions.
Judea Pearl (UCLA)
The do-calculus consists of three rules that permit the removal or insertion of observations and interventions under specific graphical conditions.
Highlights the three rules and graphical conditions for validity.
RELATED TERMS
Structural Causal Model Directed Acyclic Graph Backdoor Criterion Front-Door Criterion Causal Graph Interventional Distribution g-Formula Treatment Effect Identification Counterfactual
COMPETING TERMS
KNOWLEDGE AREAS
TARGET AUDIENCE
Data Scientist AI Engineer Information Analyst
EXTERNAL SOURCES
- Wikipedia - Do-calculus
- UCLA Causal Research - Pearl’s exposition of do-calculus
- UCLA Technical Report - Formal treatment of do-calculus rules
- Brady Neal - Intro textbook with do-calculus overview
TOP QUESTIONS
What problem does Do-Calculus solve?
It determines whether causal effects are identifiable from observational data, and how to express them using available distributions and the assumed causal graph.
Category: Definition
How many rules are in Do-Calculus?
Three rules govern insertion or deletion of observations and interventions under graphical conditions.
Category: Overview
How does Do-Calculus relate to DAGs?
It uses DAG-based separation conditions to justify transformations, relying on the SCM/DAG to encode assumptions.
Category: Example
Is Do-Calculus always necessary?
No. Simpler criteria like backdoor or front-door often suffice; do-calculus generalizes and formalizes these cases.
Category: Opinion
What is a risk when applying Do-Calculus?
Using an incorrect causal graph invalidates inferences; correctness hinges on well-justified structural assumptions.
Category: Risks
Can Do-Calculus be automated?
Yes. Tools implement identification algorithms that apply do-calculus mechanically to derive estimands.
Category: Practicality
How does Do-Calculus compare to Potential Outcomes?
Both aim at identification; do-calculus uses graphical rules, while the potential outcomes approach emphasizes design and conditional ignorability.
Category: Alternatives
Where to learn Do-Calculus?
Pearl’s texts and modern intros like Brady Neal’s book present the rules with worked examples.
Category: Learning
What are common mistakes with Do-Calculus?
Misreading d-separation or adjusting for colliders; careful graph analysis is crucial.
Category: Mistakes