causallink

A causallink is a directed connection in a causal graph or model that encodes a causal influence of one variable on another. In this context, nodes represent variables or factors, while causallinks (edges) denote the mechanism by which a cause affects an effect. Causallinks are central to causal inference, providing a visual and mathematical representation of assumed or estimated dependencies needed to reason about interventions and counterfactuals.

In formal terms, a causallink from X to Y suggests that changing X, holding other factors constant, will, through a specified mechanism, change the distribution or value of Y. In structural causal models, such links are described by structural equations or by conditional distributions that specify how Y depends on its direct causes, including X. Causallinks can be annotated with functional forms, probabilistic relations, or parameters that quantify the strength and nature of the influence.

Representations and analysis often involve directed acyclic graphs (DAGs) or more general causal graphs. These tools help identify confounding, mediation, and pathways through which effects propagate. Estimation of causal effects typically relies on the structure of causallinks to determine identifiability under interventions, using methods such as do-calculus, propensity scores, or structural equation modeling.

Causallinks are used across disciplines, including epidemiology, economics, social sciences, and machine learning, to reason about what would happen under interventions and to distinguish correlation from causation. Limitations arise from model misspecification, hidden confounders, and the assumptions required for identifiability.