The presence of bounded-noise in identification data implies uncertainty in the parameters of system models. Earlier results provide analytical means for determining optimal ellipsoidal regions that bound the vector of parameters. Parameter space information however is only of secondary interest in frequency response design where the main preoccupation is with the bounding of Nyquist diagrams. The present paper considers the projection of ellipsoidal regions from parameter space to the Nyquist plane and obtains results which are optimal from a frequency response point of view. As an illustration of the potential use for frequency response bounding regions, consideration is given to the problem of multivariable robust frequency response design and a numerical example is discussed.
