AI 용어집
인공지능 완전 사전
Mediator
Intermediate variable that transmits the effect of an independent variable to a dependent variable, representing the underlying mechanism by which the cause produces its effect.
Indirect Effect
Portion of the total effect of an independent variable on a dependent variable that passes through the mediator(s), quantifying the mediated influence in the causal model.
Direct Effect
Effect of an independent variable on a dependent variable that is not transmitted by mediators included in the model, representing the direct causal influence or unobserved mechanisms.
Total Effect
Sum of the direct effect and the indirect effect, representing the overall influence of an independent variable on a dependent variable, including all possible causal pathways.
Sobel Test
Statistical test used to assess the significance of the indirect mediation effect by calculating a ratio of the indirect effect to its standard error, assuming a normal distribution.
Bootstrap in Mediation
Non-parametric resampling method for estimating confidence intervals of the mediation effect without assuming a normal distribution, particularly useful for small samples.
Multiple Mediation
Mediation model involving multiple mediators simultaneously, which can be parallel (independent) or sequential (chained), allowing analysis of complex mechanisms.
Moderator Variable
Variable that modifies the strength or direction of the relationship between two other variables, distinguished from a mediator which explains the mechanism of this relationship.
Mediation Confounder
Variable that influences both the mediator and the dependent variable, potentially biasing the estimation of the mediation effect if not properly controlled in the analysis.
Directed Acyclic Graph
Graphical representation of causal relationships between variables where arrows indicate the direction of influence and the absence of cycles ensures the consistency of the causal model.
Sequential Mediation
Mediation analysis involving a chain of mediators where each mediator influences the next, allowing examination of complex causal mechanisms and stepwise processes.
Natural Direct Effect
Direct causal effect that keeps the mediator at its natural value under the reference treatment, distinguishing counterfactual interpretations of direct effects.
Natural Indirect Effect
Portion of the total causal effect that operates through the change in the mediator induced by the treatment, while holding other system characteristics constant.
Mediation Sensitivity
Analysis evaluating how mediation effect estimates would change in the presence of unmeasured confounders, quantifying the robustness of mediation conclusions.
Mediation with Interaction
Mediation model where the effect of treatment on the outcome depends on the level of the mediator, capturing non-linear and synergistic causal mechanisms.
Total Effect Decomposition
Analytical procedure dividing the total causal effect into direct and indirect components, sometimes including interaction effects, for a fine-grained understanding of mechanisms.
Hypothèse d'Exclusion
Postulat en médiation stipulant que le traitement n'affecte le résultat qu'à travers les médiateurs inclus dans le modèle, condition cruciale pour l'identification causale.