ericstrobl/CCI

CCI is an algorithm which discovers causation from observational data. More specifically, it is a constraint-based algorithm for causal discovery with cycles, latent variables and/or selection bias. Running CCI feels like running FCI, but CCI can handle cycles. CCI discovers a partially oriented maximal almost ancestral graph (MAAG) of some directed graph G, provided that the global Markov property and d-separation faithfulness holds according to G. Such properties are reasonable when G is the directed graph of a structural equation model with independent errors and linearity holds, for example.