1.5.1. dodiscover.constraint.ConditioningSetSelection#

class dodiscover.constraint.ConditioningSetSelection(value)[source]#

Available methods for selecting the conditioning sets when learning a skeleton.

Given a pair of nodes in a graph, (X, Y), this enumeration selects a strategy for choosing conditioning sets to be checked for conditional independence.


Allows ‘contains’ checks because of the metaclass. For example, one can run "complete" in ConditioningSetSelection, which would return True.

COMPLETE = 'complete'#

Considers all possible combinations of nodes in the graph that are not (X,Y).

NBRS = 'neighbors'#

Considers all current neighbors of (X,Y) in the graph.

NBRS_PATH = 'neighbors_path'#

Considers all neighbors of (X,Y) in the graph that are on a simple path between the two nodes.

PDS = 'pds'#

Considers all potentially d-separating sets for (X,Y) in the graph. This requires some initial collider information in the graph.

PDS_PATH = 'pds_path'#

Considers all PDS sets for (X,Y) in the graph that lie on a path between the two nodes.