scientific article; zbMATH DE number 1449647
The authors consider a nonlinear regression model \( y_i=f(X_i,w)+\varepsilon_i, \) where \(w\) is a vector of estimated parameters and \(\varepsilon_i\) is a residual error. They propose a version of an approximate nonlinear Durbin-Watson test for the hypothesis \(H_0\): \(\varepsilon_i\) are i.i.d. zero mean normal vs. \(H_1\): \(\varepsilon_t=\rho\varepsilon_{t-1}+\nu_t,\) where \(\rho\not=0\), \(|\rho|<1\), \(\nu_t\) are i.i.d. zero mean normal. The test is based on the estimation of \(w\) using a regularized least squares estimator and some linear approximation. This approximation is not described properly. The authors refer only to \textit{J. Moody} [J. Moody, S. Hanson and R. Lippmann (eds), Advances in Neural Network Processing Systems, 4, 847-854 (1992)]. For the entire collection see [Zbl 0929.00046].
- scientific article; zbMATH DE number 5233727
- The application of the durbin-watson test to the dynamic regression model under normal and non-normal errors
- A nonparametric test for serial independence of regression errors
- Testing for Autocorrelated Disturbances in Nonlinear Regression Analysis
- The Class of Models for Which the Durbin-Watson Test is Locally Optimal
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4955680)