抄録
In this paper, for the purpose of evaluating the degradation level of the performance in application with the estimated parametric model in which the estimated parameters have the disturbance, we consider the relation between the robustness of parameter estimation and the model order. In this paper, the robustness of the estimated parameter is defined as the degradation level of the performance. After the application performance cost function is derived, the gradient at the estimated parameter is focused to evaluate the disturbance level of the performance in the application. As a result, when identifying the model with the true structure by maximum-likelihood method, the disturbance level is simply given as the function of the model order. And it is shown that the larger the model order is, the more the estimated parameter lacks the robustness in the application. In addition, the relation between the robustness and a information criterion is introduced.
