A common generalization framework for two classical nonlinear models

The Threshold Autoregressive (TAR) and Exponential Autoregressive (EXPAR) models are to classical nonlinear signal models whose structure allows to reproduce phenomena such as limit cycles and chaotic regimes. In both models however nonlinearity in the parameters imposes simplifying assumptions which reduce their range of application. In this paper, we propose a common generalization framework which establishes the close connection between the two models and alleviates their limitations.