Tests of Disorder of Regression: Asymptotic Comparison

From MaRDI portal





Let the regression models \(y_ k=x_ k^ t\beta_ k+u_ k (k=1,...,n)\) be given, where \(x_ k\) are known s-dimensional vectors, \(\beta_ k\) are s-dimensional parameters and \(u_ k\) are independent \(N(0,\sigma^ 2_ k)\) variables. The hypothesis \(H_ 0:\{\beta_ 1=...=\beta_ n\), \(\sigma_ 1=...=\sigma_ n\}\) is to be tested against \(H_ 1:\{there\) exists \(k_ 0\) such that \(\beta_ 1=...=\beta_{k_ 0}\neq \beta_{k_ 0+1}=...=\beta_ n\), \(\sigma_ 1=...=\sigma_ n\}.\) The likelihood ratio tests and the tests constructed on the recursive residuals are examined using an asymptotic comparison of Bahadur type. A procedure based on a randomization of the moments of observation is suggested, which is in certain cases asymptotically better than those previously studied.











This page was built for publication: Tests of Disorder of Regression: Asymptotic Comparison

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3317919)