Table 3 Structural features that indicate different dependence types between the variables

Pairwise features

Direct causal relevance

DCR(X,Y)

There is an edge between X and Y

Transitive causal relevance

TCR(X,Y)

There is directed path between X and Y

Confounded relevance

ConfR(X,Y)

X and Y have common ancestor

Association

A(X,Y)

DCR or TCR or ConfR

Pure interactionist relevance

PIR(X,Y)

X and Y have common child

Strong relevance

SR(X,Y)

PIR or DCR

Relevance of variable sets

Strong relevance

MBS(Y)

The set consisting of Y’s parents, its children, and the other parents of its children (the Markov Blanket Set of Y)

Interaction models of relevant variables

Strong relevance

MBG(Y)

The subgraph that includes the nodes in the MBS and the incoming edges into Y and into its children (the Markov Blanket Subgraph of Y)