Augment a base DAG with edges from a Knowledge Graph — causal_augment

Takes a dagitty DAG expressing prior / expert structural knowledge (e.g. the CLORPT or Cerrado DAGs shipped with edaphos) and unions it with the subset of Knowledge-Graph edges whose confidence is at least min_confidence. Duplicate edges are dropped silently, and any edge whose insertion would introduce a directed cycle is rejected (warned but not inserted), so the returned DAG is guaranteed to be acyclic and usable for causal_adjustment_set().

Usage

causal_augment_dag(base_dag, kg, min_confidence = 0.7, allow_new_nodes = TRUE)

Arguments

base_dag: A dagitty DAG (e.g. causal_cerrado_dag()).
kg: An edaphos_causal_kg (from causal_kg_new() / LLM ingestion).
min_confidence: Numeric in [0, 1]; KG edges strictly below this threshold are ignored.
allow_new_nodes: Logical. When TRUE (default) nodes appearing only in the KG are added to the augmented DAG. When FALSE KG edges touching unseen nodes are dropped — useful when the analyst wants to lock the structural vocabulary.

Value

An object of class dagitty, augmented.