On Relations Between Weighted Least-Squares Estimators of Parametric Functions under a General Partitioned Linear Model and Its Small Models

Changli Lu,Yuqin Sun,Yongge Tian

DOI: https://doi.org/10.1007/s00184-012-0412-x

2008-01-01

Annals of the Institute of Statistical Mathematics

Abstract:We study relations between the weighted least-squares estimators (WLSEs) of given parametric functions \(\mathbf{K}_1\varvec{\beta }_1 + \mathbf{K}_2\varvec{\beta }_2\) under a general partitioned linear model \({\fancyscript{M}}=\{ \mathbf{y}, \, \mathbf{X}_1\varvec{\beta }_1 + \mathbf{X}_2\varvec{\beta }_2, \, \sigma ^2\varvec{\Sigma }\}\) and the WLSEs of \(\mathbf{K}_1\varvec{\beta }_1\) and \(\mathbf{K}_2\varvec{\beta }_2\) under the two small models \(\{ \mathbf{y}, \, \mathbf{X}_1\varvec{\beta }_1, \, \sigma ^2\varvec{\Sigma }\}\) and \(\{ \mathbf{y}, \, \mathbf{X}_2\varvec{\beta }_2, \,\sigma ^2\varvec{\Sigma }\}\). Some consequences on additive decomposition of the best unbiased linear estimator (BLUE) of \(\mathbf{K}_1\varvec{\beta }_1 + \mathbf{K}_2\varvec{\beta }_2\) under \({\fancyscript{M}}\) are also given.