Assumptions

nNo missing data or hidden variables

nLet G be a causal graph then:

nCausal Markov Assumption

nData is generated according a Bayesian Network with structure G

nE.g., Can’t have G=(X
Y) and X dependent on Y

nFaithfulness Assumption

nNo probabilistic independencies other than those represented by G

nE.g., Can’t have G=(X®Y) and X independent of Y