Coverage for src/causalspyne/utils_causallearn_g2ancestral.py: 0%

1""" 

2A graph is simpler than simplicial complex in that it only characterize 

3pairwise relationship, which enables us to project a complicated graph in 

4causallearn to ancestral ADMG (i.e. ancestral graph with only directed edges) 

5 

6We of course lose information via projection. 

7 

8This module offers another function to extract causal order from 

9causal-learn 

10""" 

11from causallearn.graph.NodeType import NodeType 

12 

13 

14def project_causallearn_g2ancestral_admg(mat_graph_causallearn): 

15 """ 

16 A general graph in causal-learn is not necessarily a PAG (it does have 

17 an attribute to denote whether it is a PAG or not) 

18 When it is a PAG, like the return from FCI: 

19 https://causal-learn.readthedocs.io/en/latest/search_methods_index/Constraint-based%20causal%20discovery%20methods/FCI.html#usage 

20 g.graph is a PAG 

21 A->B: G[A, B]=-1, G[B, A]=1, to our format, G[A, B]=0, G[B, A]=1 

22 A.->B: G[A, B]=2, G[B, A]=1, B is not an ancestor of A, 

23 to our format, G[A, B]=0, G[B, A]=1 

24 A.-.B: G[A, B]=2, G[B,A]=2, no set d-separates A and B, ????? 

25 A<->B: there is common latent cause, G[A, B]=1, G[B, A]=1, 

26 no need to convert 

27 """ 

28 for i in range(len(mat_graph_causallearn)): 

29 for j in range(len(mat_graph_causallearn[i])): 

30 if mat_graph_causallearn[i][j] != 1: 

31 mat_graph_causallearn[i][j] = 0 

32 return mat_graph_causallearn 

33 

34 

35def get_causalearn_order(g_causal_learn, node_names=None): 

36 """ 

37 get causal order from causal learn graph using api like 

38 'node.get_node_type() == NodeType.LATENT:' 

39 """ 

40 real_name_order = [] 

41 real_na_order_latent = [] 

42 nodes_order = g_causal_learn.get_causal_ordering() 

43 for node in nodes_order: 

44 if node.get_node_type() == NodeType.LATENT: 

45 # GIN-LINLAM 

46 real_na_order_latent.append(node.get_node_type) 

47 continue 

48 fake_name = node.get_name() 

49 index = int(fake_name.removeprefix("X")) - 1 

50 real_name = node_names[index] 

51 print(f"{real_name}") 

52 real_name_order.append(real_name) 

53 real_na_order_latent.append(real_name) 

54 print(f"{real_name_order}") 

55 return real_name_order, real_na_order_latent