Uncertainty models play a central role in the robust control framework. The uncertainty models and their structure determine the design trade off between performance and robustness of the closed-loop system. Therefore, given a nominal multivariable model of the system, a set of multivariable frequency response measurements, and model structure, it is desirable to generate a correspondingmodel set which tightly over bounds the given data.We show that computation of themodel set for a given structure, which is consistent with the data, can be formulated as a linear matrix inequality feasibility problem. Formulas are derived which allow comparison between model structures to assess the relative size of the each model set. The proposed algorithms are applied to the lateral-directional axis of a radio-controlled aircraft developed by NASA Langley researchers.