A factored form of the instrumental variable algorithm

A UDV^T-factored form of the instrumental variable (IV) identification algorithm is developed for identification of systems with coloured noise. The algorithm uses a UDV^T factorization technique to reformulate the conventional instrumental variable method and simultaneously produces the parameter estimates and loss functions for all models from order one to a user-specified number n with approximately the same computational effort as the conventional IV algorithm. The algorithm possesses a unique compact structure that facilitates the interpretation and implementation of IV-type algorithms. The UDV^T factorization technique used in this algorithm also results in better numerical performance than the conventional instrumental variable method.