A K-S type test of linearity for a class of time series models (Q1814704)

Consider the nonlinear autoregressive model (1) \(x_t= \varphi x_{t-1}+ \varepsilon_t h(x_{t-1}, \theta)\) with the assumptions: (A1) \(|\varphi|< 1\), \(\theta= (\theta_0, \theta_1)\in \Theta\), \(\Theta\) is an open set in \(\mathbb{R}^2\); (A2) \(\{\varepsilon_t\}\) is a sequence of independent identically distributed random variables such that \[ E\varepsilon_t= 0, \quad E\varepsilon_t^2= 1,\quad E|\varepsilon_t |^{4+\delta}< \infty, \quad\text{for some }\delta>0; \] and \(\varepsilon_t\) is independent of \(x_{t-1}\); (A3) \(h(\cdot)\) is an everywhere positive measurable function satisfying that as \(| x|\to \infty\), \(h(x)\to \infty\), \(h(x)/| x|\to 0\), and for each \(C>0\), \(\sup_{| x|\leq C} h(x)< \infty\); (A4) \(\{x_t\}\) is a \(\varphi\)-mixing sequence satisfying \[ \sum_{k=1}^\infty k(\varphi_k)^{\frac {2+\delta}{4+\delta}}< \infty, \] where \(\varphi_k= \sup_{A\in {\mathcal F}_{-\infty}^0,\;B\in{\mathcal F}_k^\infty} | P(AB)- P(A)P(B)|/ P(A)\), \({\mathcal F}_{-\infty}^0= \sigma\{x_s\), \(s\leq 0\}\), \({\mathcal F}_k^\infty= \sigma\{x_s\), \(s\geq k\}\), and \(x_t\) has continuous marginal distribution. The purpose of the note is to discuss the test of linearity for model (1).