scientific article; zbMATH DE number 3842967
The author considers the model \(X_ i=\xi_ i+\epsilon_ i\), \(i=1,2,...,n\), where the \(\xi_ i's\) and the \(\epsilon_ i's\) are unobservable independent random p-vectors and where the \(\xi_ i's\) are subject to the relation \(\alpha +\beta^ t\xi_ i=0\) for \(i=1,2,...,n\) with \(\beta^ t\beta =1\). Assuming that the \(\xi_ i's (\epsilon_ i's)\) are \(N(\mu,\Sigma_{\xi}) (N(0,\delta^ 2I))\) distributed and that \(\delta^ 2\) is known, \textit{R. L. Brown} and \textit{F. Fereday} [Biometrika 45, 136-153 (1958; Zbl 0088.355)] constructed a confidence region for \((\alpha,\beta)\) by observing that \[ Q = \sum^{n}_{i=1} (\alpha+\beta^ t X_ i)^2 / \delta^2 \] follows the chi-squared distribution with n degrees of freedom. The asymptotic distribution of Q is given when \(\delta^ 2\) is replaced by its maximum likelihood estimator, the smallest eigenvalue of the sample covariance matrix. Using this result a method for constructing a conservative confidence region is proposed which turns out to be inferior to the maximum likelihood method presented by the author and \textit{Y. Isogawa}, J. Jap. Stat. Soc. 11, 119- 126 (1981; Zbl 0496.62050).
- scientific article; zbMATH DE number 3862246
- Asymptotic variance-covariance matrices for the linear structural model
- Asymptotic theory for partly linear models
- Asymptotic theory for estimation of error distribution in linear model
- Asymptotic theory of least squares estimator of a particular nonlinear regression model
- Asymptotic theory in heteroscedastic nonlinear models
- Asymptotic theory of nonlinear regression
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 Q3313100)