Linear approximation of the Threshold AutoRegressive model: an application to order estimation

Abstract This paper proposes a linear approximation of the nonlinear Threshold AutoRegressive model. It is shown that there relation between autoregressive order threshold model and its moving average approximation. The main advantage this can be found in extension some theoretical results developed setting to domain. Among them proposed new estimation procedure for models whose performance compared, through Monte Carlo study, other criteria largely employed context.