For nonparametric causal models represented by acyclic directed mixed graphs, every identifiable interventional distribution can be transformed into an expression involving only the observational distribution by the ID algorithm, and the transformations used by the algorithm are derivable using do-calculus together with ordinary probability rules.