Some comments on specification tests in nonparametric absolutely regular processes

From MaRDI portal





Let \((X_{t},Y_{t})\) denote a strictly stationary discrete time stochastic process with \(Y_{t}\in R, X_{t}\in R^{d}\) and \(Y_{t}=m(X_{t})+\varepsilon_{t}\), where \(\varepsilon_{t}\) is an unobservable error with mean \(0\) and \(m\) is an unknown function. Let \(g_1,\ldots,g_{p}\) denote linearly independent given regression functions. The authors are interested in testing the hypothesis of linearity \[ H_0:\;m(x)=\sum_{j=1}^{p}\theta_{j}g_{j}(x),\quad \forall x\in \text{supp}(\pi) \text{ for some }\theta_1,\ldots,\theta_{p}\in R, \] where \(\pi\) is the density of the stationary distribution of \(X_{i}\). Let \(g(x)\) be the vector of regression functions and \(\widehat\theta_{n}\) denote the least squares estimator of the unknown parameter \(\theta\). The authors define the test statistic for the hypothesis of linearity by \[ T_{n}=(n(n-1))^{-1} \sum_{i\neq j}K_{h}(X_{i} -X_{j})\widehat\varepsilon_{i}\widehat\varepsilon_{j}, \] where \(\widehat\varepsilon_{i}=Y_{i}-\widehat\theta_{n}^{T}g(X_{i})\) are the residuals obtained under the assumption of linearity; \(K\) is a kernel function, and \(K_{h}(x)=h^{-d}K(x/h)\). The asymptotic behaviour of \(T_{n}\) is studied. The asymptotic normality of \(U\)-statistics is proved under fixed alternatives with a different rate of convergence as under the null hypothesis. It is demonstrated that several results regarding the asymptotic distribution of goodness-of-fit tests are incorrectly stated in the literature. The authors show that results regarding asymptotic equivalence of a nonparametric autoregression and a nonparametric regression model are not applicable for the asymptotic analysis of goodness-of-fit tests under fixed alternatives.





Describes a project that uses

Uses Software






This page was built for publication: Some comments on specification tests in nonparametric absolutely regular processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4828175)