Discussion Papers, Interdisciplinary Research Project 373: Quantification and Simulation of Economic Processes 2001,101
This paper suggests a general functional-coefficient regression model in a form of ARX time series model. Contrast to the common threshold variable in the previous works, our model allows each coefficient to possess a different threshold variable and can cover a wide range of nonlinear dynamic processes. The estimation procedure consists of two steps; local linear smoothing and marginal integration. The asymptotic normality of the proposed estimator is derived with the explicit form of bias and variance.