Functional coefficient autoregressive models: estimation and tests of hypotheses (Q2740102)

Functional coefficient autoregressive (FAR) models of the form NEWLINE\[NEWLINEX_t=f_1(X_{t-d})X_{t-1}+\cdots +f_p(X_{t-d})X_{t-p}+\varepsilon_tNEWLINE\]NEWLINE are considered. The authors propose a local linear estimator for estimating the coefficient functions nonparametrically. Testing procedures to detect if the coefficient functions are constant functions and if there are any discontinuous points in coefficient functions are developed. Simulation studies and the analysis of an example are given.