A New Approach to the Threshold Autoregressive Models

Time series analyzing is very important tool for economic and financial system. However, recent developments show that financial systems are known in a structural change. Therefore, nonlinear time series have been analyzed for past decades because of these changes. In this paper, we consider Threshold Autoregressive (TAR) model. The most popular method for estimating the parameters and threshold value is least square (LS) method. However, LS method is not robust to the outliers and departures from normality. Therefore, we propose a robust version of estimation in order to provide robust results. JEL classification numbers: C01, C22, C13.