A note on minque for normal models
- 1 January 1976
- journal article
- research article
- Published by Taylor & Francis in Mathematische Operationsforschung und Statistik
- Vol. 7 (5) , 707-714
- https://doi.org/10.1080/02331887608801330
Abstract
Consider a normal linear model where e has covariance matrix with c = (c1 … cp) ′ unknown. Restricting estimators to the class of functions for all β we prove, that Minque is locally minimum variance unbiased estimator. This result is extended to the class of all unbiased estimators if the range of X is assumed to be an invariant subspace with respect to all linear mappings given by V 1 … Vp . Furthermore necessary and sufficient conditions for the existence of uniformly best unbiased estimators (BUE) or only uniformly best invariant unbiased estimators (BIUE) are developed. They are small extensions of earlier results published by C.R. Rao [7] and J. Seely [8]. A sufficient condition for a BUE t exist given in 8, is shown to be not necessary.Keywords
This publication has 6 references indexed in Scilit:
- Best quadratic unbiased estimators for variance components in the balanced two-way classification modelMathematische Operationsforschung und Statistik, 1975
- Estimability of parameters of the covarlance matrix and variance componentsMathematische Operationsforschung und Statistik, 1974
- Bates and best quadratic unbiased estimators for parameters of the covariance matrix in a normal linear modelMathematische Operationsforschung und Statistik, 1974
- Minimum variance quadratic unbiased estimation of variance componentsJournal of Multivariate Analysis, 1971
- Estimation of variance and covariance components—MINQUE theoryJournal of Multivariate Analysis, 1971
- Quadratic Subspaces and CompletenessThe Annals of Mathematical Statistics, 1971