PAG Edge Types
Edge Types Present Relationships Absent Relationships
A --> B A is a cause of B.
It may be a direct or indirect cause that may include other measured variables. Also, there may be an unmeasured confounder of A and B. 		B is not a cause of A
A <-> B There is an unmeasured confounder (call it L) of A and B. There may be measured variables along the causal pathway from L to A or from L to B. A is not a cause of B.
B is not a cause of A.
A o-> B Either A is a cause of B (i.e, A --> B) or there is an unmeasured confounder of A and B (i.e, A <-> B) or both. B is not a cause of A.
A o-o B Exactly one of the following holds:
  1. A is a cause of B
  2. B is a cause of A
  3. there is an unmeasured confounder of A and B
  4. both a and c
  5. both b and c
PAG Edge Specialization Markups: If the graph is a PAG and PAG edge-specialization markup is turned on, then the following are also true of directed edges in the graph. (1) If a directed edge is solid, that means there is no latent confounder for that directed edge (i.e., the edge is visible, which means that for linear models its coefficient can be estimated); (2) If a directed edge is dashed, there is possibly a latent confounder (so that its coefficient *may not* be estimable). In addition, (3) If a directed edge is thickened, that means the edge is definitely direct (which means that the directed edge appears in the true DAG). (4) Otherwise, if the directed edges is not thickened, then it is edge is *possibly direct* (which means the directed edge may or may not appear in the true DAG--there may be an indirected directed path instead).
Also, the above breakdown of types assume there is no selection bias. Selection bias occurs when there is a common latent child of two measured nodes, thus: X→(L)←Y and is indicated in an PAG as X—Y.