Collider

As explained in Causal AI

A collider is a variable in a causal graph with directed paths entering it from at least two other variables, so those paths converge at that variable.

Suppose side-quest engagement affects in-game purchases, and an analyst wants to estimate that effect. Total inventory may look like a useful predictor, but it is caused by both purchases and items won. In the causal graph, it is a collider. Adding it to an adjustment set can introduce collider bias rather than improve the estimate. section 11.2 section 11.4.1

A collider is a node where arrows on a path converge: X → M ← Y. Unlike a chain, X → M → Y, or a common cause, X ← M → Y, this path is blocked by default. Conditioning on M—or on a descendant of M—can open it. Whether two variables are d-separated depends on every path between them, not just this one. section 4.2.1 section 4.2.2

Wet grass

A sprinkler and rain can each make grass wet. Assume sprinkler use and rain are independent: knowing the sprinkler is off does not, by itself, tell you whether it rained. Once you know the grass is wet, however, learning that the sprinkler is off makes rain more plausible. Conditioning on their shared effect creates dependence between its causes. section 4.1.1

Figure 4.12 This path from u to v is d-separated by default because it contains a collider m. The collider is analogous to an open circuit. Blocking on m or any of its descendants d-connects the path and figuratively closes the circuit.

Compare the default blocked collider path with the opened path after conditioning on the collider or a descendant. section 4.2.1

Check a proposed adjustment set against the whole graph. A node can block one path yet act as a collider on another: conditioning on M blocks U ← M → V but opens U ← I → M ← J → V. Blocking the newly opened path also requires I or J, yielding minimal d-separating sets {I, M} or {J, M} in this example. Controlling for every available covariate is therefore not a safe default for causal effect estimation. section 4.2.2 section 11.4.1

Colliders also help orient a causal graph. If two variables appear independent but become dependent when conditioning on a third, that pattern can support arrows into the third variable. Such conclusions rely on the graph’s causal and statistical assumptions; dependence alone generally does not determine an edge’s direction. section 4.6.3

How Causal AI covers it

How deep it goes

Defines the graph motif, works through conditional dependence and multi-path blocking, and connects the rule to estimation; does not derive a numerical bias formula.

Examples

Uses wet grass, inherited blood types, multi-path DAGs, and total inventory in an online gaming causal graph.

What it assumes

Assumes familiarity with causal DAGs and conditioning while defining collider paths and their d-separation behavior explicitly.

What you can do after

Diagnose/debug with it: spot adjustment choices that open a path and recognize conditional-dependence evidence for arrow direction.

How it connects

Links conditional independence and d-separation to Markov equivalence, causal discovery, adjustment sets, and collider bias.

liveBookmanning.com

Read Causal AI on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Causal Inference for Data ScienceA collider is a variable that is causally influenced by two other variables and whose conditioning can induce or alter their association.

Compare how 2 books explain collider →

Related concepts

• conditioningused by · Conditioning is restricting or guiding a probabilistic or generative process using specified information, variables, values, or events.
• collider biasproduces · Collider bias is bias caused by conditioning on a collider or its descendant, thereby inducing an association between its causes.
• d-separationcontrasts with · D-separation is a graphical criterion for determining whether a conditioning set blocks all relevant paths between variables in a graph, implying conditional independence under the graph’s assumptions.
• correlationcontrasts with · Correlation is a statistical relationship describing how two or more variables or quantities vary together, including the strength and direction of their association, without by itself implying causation.
• causal discoveryaddresses · Causal discovery is the process of inferring causal relationships or a causal graph from data under assumptions about the data-generating process.

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