AI-ordlista
Den kompletta ordlistan över AI
Directed Acyclic Graph (DAG)
Graphical representation of causal relationships between variables where nodes represent variables and directed edges indicate direct causal influence without possible cycles.
Average Treatment Effect (ATE)
Expected average difference between potential outcomes with and without treatment across the entire population, fundamental measure of causal impact in intervention evaluation.
Mediation Analysis
Causal method decomposing the total effect of a treatment into direct effect and indirect effect through intermediate variables (mediators) to understand mechanisms of action.
Rubin Causal Model
Theoretical framework based on potential outcomes where each unit has counterfactual outcomes for each treatment state, foundation of modern causal inference.
Regression Discontinuity Method
Quasi-experimental design exploiting eligibility thresholds to estimate local causal effects by comparing units just above and below the cutoff point.
Causal Score
Function summarizing the information necessary for confounding bias adjustment, generalization of propensity score including information about causal relationships between variables.
Pearl's Causality
Causality approach based on directed acyclic graphs and do-calculus, allowing formal representation of causal knowledge and counterfactual reasoning.
Conditional Average Treatment Effect (CATE)
Average causal effect conditioned on specific unit characteristics, allowing identification of heterogeneities in treatment effects to personalize interventions.
Front-door criterion
Causal identification strategy using an observable mediator that blocks all paths between treatment and outcome, allowing causal estimation even in the presence of unmeasured confounding.
Randomization test
Experimental validation of causal relationships through random allocation of treatment, systematically eliminating confounding biases and providing the most robust causal evidence.