Non-linear "Tuning" of Frankel's Exchange Rate Model

In this essay I've demonstrated that there is evidence of unstable and non-linear relationship between fundamental variables and exchange rates. I have tried to "tune" Frankel's (1979) real interest differential model of exchange rate fluctuation. I have distinguished between Czech crown/Euro market and Euro/U.S. Dollar market because there is a different behaviour of market participants (FX dealers). Questionnaire surveys indicate that the interest rates play a ominant role in professionals' decision-making process on the UR/USD market. In that aspect I have extended Frankel's (1979) original RID model by allowing the constant and the coefficients for the short and long term interest rates to switch, depending on the value of the state variable. It has improved the fit of Frankel's model. On the CZK/EUR market, questionnaire surveys indicated that trade balance plays a dominant role in FX dealer's decision-making process. I suggested extending the model with trade balance variable as an approximation for income. Unfortunately there are no good time-series available for to strongly authenticate this idea in a way of non-linearity. But generally speaking, it seems that non-linear modelling of exchange rates is able to improve our understanding of exchange rate determination.