Descendant

As explained in Causal Inference for Data Science

A descendant is a variable reachable from another variable by following one or more directed causal arrows.

A descendant of a node X is any node reachable through a causal path beginning at X: X → … → Z. In the graph X → M → Z1 ← C → Z2 ← Y, both M and Z1 are descendants of X, while Z2 is not. section 7.1.6

Descendants matter because they lie downstream of a treatment or cause. They cannot be confounders, and conditioning on them can block part of X’s causal effect. A descendant of a collider can also transmit information about that collider and open an otherwise closed path. section 7.1.6 section 7.2

Use descendant status as a practical adjustment check:

• Exclude descendants of X when selecting variables for the back-door criterion; otherwise, adjustment may remove causal paths. section 7.2
• Treat descendants of colliders cautiously: conditioning on one may open a path between variables that was previously blocked. section 7.1.6
• Do not treat every post-treatment variable as a valid control; conditioning on it may induce selection bias or reduce the estimated causal effect. section 7.3.3

Treatment and return to work

Suppose X is a drug treatment, Y is recovery time, and Z is the number of days until a patient returns to work. If Z is downstream of treatment, analyzing only patients who returned quickly can make the drug appear more effective than it is for the full population. section 7.3.3

Figure 7.14 Conditioning on a descendant (Z₂) of a collider (Z₁) can open new correlations (between M and N).

Trace how conditioning on Z2, a descendant of collider Z1, can open a new correlation between M and N. section 7.1.6

How Causal Inference for Data Science covers it

How deep it goes

Defines descendants formally, illustrates them in DAGs, and traces their roles in confounding, collider bias, and causal-effect estimation.

Examples

Uses DAG nodes M, Z1, and Z2, plus drug treatment, recovery time, and return-to-work examples.

What it assumes

Assumes familiarity with directed acyclic graphs, causal paths, conditioning, colliders, and confounders.

What you can do after

Diagnose adjustment risks by identifying downstream variables that could block effects, open collider paths, or induce selection bias.

How it connects

Connects descendants to causal paths, mediators, colliders, confounders, d-separation, back-door criteria, and selection bias.

liveBookmanning.com

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

Also explained in

• 

  liveBookmanning.com
  Web Design Playground, Second EditionA descendant element is an HTML element contained within another element directly or through one or more intermediate elements in the document tree.

Compare how 2 books explain descendant →

Related concepts

• focus-within pseudo-classused by · The focus-within pseudo-class matches an element when it or one of its descendants has focus.
• confoundercontrasts with · A confounder is a variable associated with both an explanatory or treatment variable and an outcome, whose influence can distort their apparent relationship or estimated causal effect.
• backdoor criterionused by · The backdoor criterion is a graphical condition for identifying a causal effect by requiring a variable set to block every backdoor path while excluding treatment descendants.
• direct descendant combinatorcontrasts with · The CSS child combinator, written as >, matches elements that are direct children of a specified parent element.
• ancestorpart of · An ancestor is an entity higher in a hierarchical structure, reached by repeatedly following parent relationships toward the root.

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.