A IV based criterion for model order selection (Q1923083)

The authors propose a new criterion for estimating the order of a scalar ARX model. The key-idea of the method is to replace the usual least squares loss function by a suitable estimate of its noiseless value. The weak convergence of the estimate is established with general assumptions on the noise together with relatively strong conditions on the input. However, the transfer function of the noise may be time-varying or of infinite order. Finally, the convergence is illustrated by numerical simulations.