Padé approximation and its application in time series analysis (Q1186530)

The auto-regressive-moving average (ARMA) model is discussed. A new method for specification of the order and estimation of parameters of the ARMA \((p,q)\) model is proposed. It is based on the theory of Padé approximations. First, one fits a very high order auto-regressive (AR) model to the given stationary time series using the famous Box-Jenkins technique. Then the power series with evaluated coefficients of the AR model is approximated by means of a rational function of two polynomials of degrees \(p\), \(q\). Their coefficients coincide with the coefficients of the ARMA\((p,q)\) model. At the second step the classical technique of Padé approximation is applied.