Abstract: This paper presents a novel method with adjustable parameters for the reduction of high order linear time invariant dynamic system. The method emphasizes the "quasiequivalent" relation between the reduced order model and the prototype from the dynamic characteristics view of the systems. The desirable values of the adjustable parameters are selected closely to approximate the dominant frequency response data of the prototype system. The approximants not only can retain steady state characteristics and stability, but also can selectively retain other dominant performance specifications of the prototype system (such as bandwidth and relative stability) and other characteristics over a desired frequency range. Finally, it is shown by examples that this method has established a relation between the theory of model reduction and the design of control systems.