In this paper, we consider the modelling problem for linear discrete-time systems using the following two kinds of information from the objective system:
1) values of the transfer function evaluated at finite points in the open unit disk (first-order information),
2) values of causal part of the power spectrum function evaluated at the same points as in 1) (second-order information).
The modelling problem is formulated as a partial or complete interpolation problem of the given data, and is solved based on the Nevanlinna-Pick theory. Since this modelling problem is reduced to the Mullis and Roberts problem in a special case, the results obtained here are available to their problem with a slight modification.