Closed mschauer closed 3 years ago
Essentially, the PC algorithm starts with a complete graph of all possible connections in https://github.com/mschauer/CausalInference.jl/blob/4bde12f104b1fe233fb9b3b0e065d8a8a97c85ec/src/skeleton.jl#L26
We can already remove some edges corresponding to known conditional independencies using a keyword argument.
Second, in https://github.com/mschauer/CausalInference.jl/blob/4bde12f104b1fe233fb9b3b0e065d8a8a97c85ec/src/pc.jl#L126 we start with each edge being bi-directed and then we remove directions until ending up with a directed graph. That is the point to a priori orient edges too. @archanarw
Essentially, the PC algorithm starts with a complete graph of all possible connections in https://github.com/mschauer/CausalInference.jl/blob/4bde12f104b1fe233fb9b3b0e065d8a8a97c85ec/src/skeleton.jl#L26
We can already remove some edges corresponding to known conditional independencies using a keyword argument.
Second, in https://github.com/mschauer/CausalInference.jl/blob/4bde12f104b1fe233fb9b3b0e065d8a8a97c85ec/src/pc.jl#L126 we start with each edge being bi-directed and then we remove directions until ending up with a directed graph. That is the point to a priori orient edges too. @archanarw