IdentificationofPartiallyObserved LinearCausalModels: GraphicalConditionsfortheNon-Gaussianand HeterogeneousCases