The development of non-linear unit root tests and a suggested a test

The aim of this study is to propose a new unit root test according to the needs considering the development in the literature. In recent years, there are many studies examining the nonlinear structure of time series. This allows evaluation of time series as a nonlinear. Unit root tests developed in the previous works and currently used by researchers were used under the assumption that the time series was linear. In this case, evaluating a nonlinear time series as a linear structure will make it difficult to determine its unit root structure. However, the phenomenon of structural break is nowadays described as frequency-based rather than time-based methods. As such, it is easier to identify the impact of structural break and eliminate the negative impact on unit root tests. In this case, it is seen that the required unit root structure should be defined in non-linear structure and under structural breaks. According to this need, KSS (2003), Sollis (2009) and Kruse (2011) tests were examined as modeling by non-linear structure under the exponential smooth transition autoregressive (ESTAR). FADF (2010) and FKSS (2010) tests are examined structural break modeling with using fourier transforms. The examination was carried out to determine the critical values of the tests and to determine the size and power characteristics of the finite samples. As a result of examination of these test families under structural fracture; FSollis and FKruse tests were built. In the FSollis test, asymmetric nonlinear structure was taken into consideration. However, in the FKruse test, differentiation of the threshold value was taken into consideration. Critical values of these proposed test structures, size and power characteristics of finite samples were examined. By using these tests that are suitable in terms of size and power characteristics; empirical application has been achieved with US Dollar and Euro exchange rates in the selected time period.