We overview three kinds of possibilistic graphical models (based on directed acyclic graphs) and present, how they can be expressed by means of non-graphical approach to multidimensional models, so-called compositional models. We show that any of these graphical models can be transformed into a compositional model, but not vice versa. The only exception are directed possibilistic graphs, which are as general as so-called prefect sequences of low-dimensional distributions.