{
  "id": "collider",
  "title": "Collider",
  "lang": "en",
  "source_lang": "en",
  "summary": "A collider is a node on a path in a directed graph where two arrowheads meet; conditioning on it can induce dependence between otherwise independent variables.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Collider node",
    "Common effect"
  ],
  "generated_at": "2026-11-06",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/collider/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/collider/",
      "json": "https://aiwiki.page/zh/collider.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A collider is a node on a path in a directed graph at which the two adjacent edges both have arrowheads pointing toward that node. The simplest example is \\(X \\rightarrow C \\leftarrow Y\\), where \\(C\\) is a collider. In probabilistic graphical models and causal inference, colliders are important because conditioning on a collider, or sometimes on one of its descendants, can make otherwise independent variables statistically dependent. This mechanism underlies an important form of selection bias known as collider bias."
    },
    {
      "heading": "Graphical definition",
      "text": "Consider a path containing three consecutive, distinct nodes \\(A\\), \\(B\\), and \\(D\\). The middle node \\(B\\) is a collider on that path when the local configuration is\n\n\\[A \\rightarrow B \\leftarrow D.\\]\n\nBoth arrows point into \\(B\\). In contrast, \\(B\\) is a non-collider in any of these configurations:\n\n\\[A \\rightarrow B \\rightarrow D, \\qquad A \\leftarrow B \\rightarrow D, \\qquad A \\leftarrow B \\leftarrow D.\\]\n\nA path is considered without requiring its arrows to point in the direction in which the path is traversed.\n\nCollider status is relative to a particular path, rather than an absolute property of a node. A node with several incident edges may be a collider on one path and a non-collider on another.\n\nIn a directed acyclic graph (DAG), a collider has two parents along the path under consideration. If the graph is interpreted causally, it is a common effect of those parents. A node reachable from it by following one or more arrows is called its descendant.\n\nThe configuration \\(X \\rightarrow C \\leftarrow Y\\), with no edge between \\(X\\) and \\(Y\\), is also called an unshielded collider or v-structure. A collider need not be unshielded: the two nodes pointing into it may also be adjacent."
    },
    {
      "heading": "Probability and conditional dependence",
      "text": "The distinctive behavior of a collider can be seen in a three-variable Bayesian network. The graph\n\n\\[X \\rightarrow C \\leftarrow Y\\]\n\nimplies the factorization\n\n\\[p(x,y,c)=p(x)\\,p(y)\\,p(c\\mid x,y).\\]\n\nHere \\(p\\) denotes a probability mass function or, where appropriate, a probability density function. Marginalizing over \\(C\\) gives\n\n\\[p(x,y)=p(x)\\,p(y),\\]\n\nso \\(X\\) and \\(Y\\) exhibit statistical independence.\n\nAfter conditioning on \\(C=c\\), however,\n\n\\[p(x,y\\mid c) = \\frac{p(x)\\,p(y)\\,p(c\\mid x,y)}{p(c)},\\]\n\nfor values with \\(p(c)>0\\). This generally does not factor into \\(p(x\\mid c)p(y\\mid c)\\). Knowledge of \\(C\\) can therefore make information about \\(X\\) relevant to predicting \\(Y\\).\n\nThis effect is often called explaining away. If either of two causes can produce an observed outcome, learning that one cause is present may reduce the need to attribute the outcome to the other. Nevertheless, collider structure alone does not determine the sign or strength of the induced association, and particular probability distributions may yield no induced dependence.\n\nA binary example\n\nLet \\(X\\) and \\(Y\\) be independent binary random variables, each equal to 1 with probability \\(1/2\\). Define \\(C=1\\) if at least one of \\(X\\) or \\(Y\\) equals 1.\n\nBefore conditioning,\n\n\\[P(Y=1\\mid X=0)=P(Y=1\\mid X=1)=\\frac12.\\]\n\nAmong observations with \\(C=1\\), the possible pairs are \\((1,0)\\), \\((0,1)\\), and \\((1,1)\\), each with conditional probability \\(1/3\\). Consequently,\n\n\\[P(Y=1\\mid X=0,C=1)=1,\\]\n\nwhereas\n\n\\[P(Y=1\\mid X=1,C=1)=\\frac12.\\]\n\nThe negative association is produced by restricting attention to the common effect. It does not imply that changing \\(X\\) would causally change \\(Y\\)."
    },
    {
      "heading": "Colliders and d-separation",
      "text": "The general graphical rule governing this behavior is d-separation. In a DAG, a path is active, or open, given a conditioning set \\(Z\\) when:\n\n- every non-collider on the path is outside \\(Z\\); and\n- every collider on the path is either in \\(Z\\) or has a descendant in \\(Z\\).\n\nTwo nodes are d-separated by \\(Z\\) if no path between them is active given \\(Z\\). Under the graphical Markov property, d-separation implies conditional independence.\n\nThus an unconditioned collider blocks a path, while conditioning on it can open that path. Conditioning on a descendant can also activate the collider because the descendant may convey information about it. For example, in\n\n\\[X \\rightarrow C \\leftarrow Y, \\qquad C \\rightarrow S,\\]\n\nconditioning on \\(S\\) can open the path between \\(X\\) and \\(Y\\), even when \\(C\\) is not observed.\n\nOpening one collider does not necessarily open an entire path: another collider may remain inactive, or a conditioned non-collider may block it elsewhere. Moreover, an open path permits statistical dependence but does not guarantee it for every compatible probability distribution. The additional faithfulness assumption rules out conditional independences that are not implied by d-separation."
    },
    {
      "heading": "Collider bias and selection",
      "text": "Collider bias arises when conditioning activates a path that creates a misleading association for the question being studied. Conditioning need not mean explicitly calculating a conditional probability. It can occur through:\n\n- restricting a sample to observations meeting an eligibility criterion;\n- stratifying results by a common effect;\n- including a common effect as a covariate in a regression;\n- analyzing only records whose inclusion depends on several variables.\n\nFor example, suppose two independent attributes both affect admission to a program. Among admitted applicants, a lower value of one attribute may tend to be accompanied by a higher value of the other. Admission is a collider, and restricting the analysis to admitted applicants conditions on it. The association within the selected group need not describe the applicant population.\n\nBerkson’s paradox is a classic manifestation of this mechanism: selection into an observed group can generate associations absent from the population from which that group was drawn. Selection effects may also distort or reverse associations already present.\n\nNot every analysis involving a collider is inherently mistaken. A conditional association within a selected population may itself be the intended object of description. Bias concerns a mismatch between the quantity estimated and the target population or causal effect."
    },
    {
      "heading": "Distinction from confounders and mediators",
      "text": "Colliders differ from two other common causal structures.\n\nRole of \\(Z\\) | Basic structure | Effect of conditioning on \\(Z\\)\nCommon-cause confounder | \\(X \\leftarrow Z \\rightarrow Y\\) | Blocks this common-cause path\nMediator | \\(X \\rightarrow Z \\rightarrow Y\\) | Blocks this directed causal path\nCollider | \\(X \\rightarrow Z \\leftarrow Y\\) | Opens this path, if no other node blocks it\n\nA confounder is not simply any variable associated with both exposure and outcome. In the basic common-cause structure, it precedes and influences both. Conditioning on it can remove an association transmitted along that path.\n\nA mediator transmits part of a causal influence. Conditioning on it changes the causal pathways represented in a comparison and generally does not preserve the total causal effect.\n\nA collider instead receives arrows from both neighboring nodes on the path. Adjusting for it can introduce rather than remove an association. Consequently, adding more covariates to a regression model does not necessarily improve causal estimation.\n\nThese roles remain path-dependent: one variable can block one path while opening another. The full graph and the target effect matter, not merely a variable’s label."
    },
    {
      "heading": "Role in causal discovery",
      "text": "Unshielded colliders are important in causal discovery, where conditional-independence patterns help constrain possible graph structures. The graphs\n\n\\[X \\rightarrow C \\rightarrow Y, \\qquad X \\leftarrow C \\rightarrow Y, \\qquad X \\leftarrow C \\leftarrow Y\\]\n\nshare the same d-separation pattern on these three nodes. The unshielded collider \\(X \\rightarrow C \\leftarrow Y\\) has a different pattern.\n\nTwo DAGs are Markov equivalent—meaning that they imply the same d-separation relations—exactly when they have the same underlying undirected adjacencies and the same unshielded colliders. Collider orientations therefore provide information that cannot be obtained from adjacency alone.\n\nInferring these structures from observations requires assumptions and reliable conditional-independence assessments. Finite samples, unobserved common causes, selection, and violations of faithfulness can limit what such methods establish. A correlation that appears after adjustment is not, by itself, proof of a collider."
    }
  ],
  "links": [
    "directed-graph",
    "probabilistic-graphical-model",
    "causal-inference",
    "selection-bias",
    "directed-acyclic-graph",
    "bayesian-network",
    "probability-mass-function",
    "probability-density-function",
    "statistical-independence",
    "random-variable",
    "d-separation",
    "conditional-independence",
    "conditional-probability",
    "confounding",
    "linear-regression"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}