Fork

As explained in Causal Inference for Data Science

A fork is a three-node causal structure in which one central node is a common cause of two other nodes, as in X←Z→Y.

A fork is a three-node causal pattern, X←Z→Y, in which Z is a common cause of X and Y. The shared cause can make X and Y correlated even when neither causes the other. section 7.1.4

Forks matter when separating prediction from causal reasoning. A correlation between X and Y may flow through Z, so estimating an effect requires considering whether conditioning on Z blocks this spurious path. section 7.1.4 section 7.1.6

Synthetic fork data

Generate a binary Z, then generate binary X and Y from Z independently. The resulting X and Y can be correlated overall, while conditioning on Z removes the correlation in the fork structure. section 7.1.4

Use the fork pattern to reason about paths:

• Recognize X←Z→Y as a noncausal path between X and Y. section 7.1.4
• Treat Z as a candidate variable for blocking spurious association when estimating the causal effect of X on Y. section 7.1.6
• Do not assume every fork is a confounder: an M-structure can contain a fork without Z directly influencing both target variables. section 7.1.4

Table 7.3 Pure fork DAG

DAG	Equations

Inspect the pure fork DAG and its equations to connect the three-node diagram with its factorized distribution. section 7.1.4

How Causal Inference for Data Science covers it

How deep it goes

Explains the three-node structure, conditional blocking, synthetic simulations, and the limit that forks are not always confounders.

Examples

Uses pure fork diagrams, binary synthetic datasets, conditional correlations, and an M-structure example.

What it assumes

Assumes familiarity with DAG paths and probability distributions while defining forks through simple three-node examples.

What you can do after

Diagnose spurious association by identifying common-cause paths and checking whether conditioning on the fork blocks them.

How it connects

Connects forks with DAGs, chains, colliders, confounders, conditional independence, and d-separation.

liveBookmanning.com

Read Causal Inference for Data Science on liveBook Get the book at manning.com

Related concepts

• pathkind of · A path is an ordered sequence of connected elements representing a route through a graph, structure, or space.
• conditional independenceproduces · Two variables are conditionally independent given specified information when their conditional joint distribution factors into the product of their conditional distributions.

This explainer was generated by Manning's knowledge graph from the text of Causal Inference for Data Science. The section links point to the parts of the book it draws on.