Two approaches are presented for the modeling of the impedance matrix of a random medium: one parametric and the other nonparametric. The former allows to take into account the data uncertainties while introducing a model error, that yields, in some cases, very high levels. The latter is based on a much simpler, deterministic, model, for which both data uncertainties and model errors are accounted for. When the model error is negligible, the parametric approach can be used for the identification of the parameters of the nonparametric model of the impedance matrix.