Transfer function estimation: A smoothness priors method

Abstract

A smoothness priors approach is taken to transfer function estimation from stationary time series data. An impulse response model plus an additive AR noise model each of order M is assumed. This is algebraically equivalent to an ARMAX plus white noise model. Frequency domain smoothness priors are assumed on the ARMAX polynomials and smoothness hyperparameters balance the tradeoff between the infidelity of the model to the data and the infidelity of the model to the smoothness constraints. The likelihood of hyperparameters is maximized by a constrained least squares-gradient search computational procedure. Wind velocity - differential pressure data from the Schmehausen hyperbolic cooling tower is analyzed.