Predictor input selection for the identification of dynamic models embedded in networks

Recently, the Two-Stage method has been proposed as a tool to obtain consistent estimates of modules embedded in complex dynamic networks. However, for this method the variables which must be included in the predictor model are currently not considered as a user choice. In this paper it is shown that there is considerable freedom as to which variables to include in the predictor model as inputs, and still obtain consistent estimates of the module of interest. Conditions that the choice of predictor inputs must satisfy are presented. Attention is focused on choosing the smallest number of predictor inputs. This could be an advantage if the node signals must be measured using sensors that are expensive. Efficient algorithms are presented for checking the conditions and obtaining the estimates