Best equivariant estimator of regression coefficients in a seemingly unrelated regression model with known correlation matrix
Best equivariant estimator of regression coefficients in a seemingly unrelated regression model with known correlation matrix
复制标题
DOI：
10.1007/s10463-015-0512-2
复制
发表时间：
2016-08
影响因子：
1
通讯作者：
H. Kurata;S. Matsuura
中科院分区：
文献类型：
--
作者：
H. Kurata;S. Matsuura
本文导出了具有椭圆对称误差的貌似不相关回归模型回归系数的最佳等变估计量。考虑了位置和尺度变换组的等方差。我们假设误差项的相关矩阵是已知的。由于相关矩阵是群作用下的最大不变参数，因此本文所处理的模型在参数空间上只生成一个轨道。结果表明，BEE可以看作是一个广义的最小二乘估计。
This paper derives the best equivariant estimator (BEE) of the regression coefficients of a seemingly unrelated regression model with an elliptically symmetric error. Equivariance with respect to the group of location and scale transformations is considered. We assume that the correlation matrix of the error term is known. Since the correlation matrix is a maximal invariant parameter under the group action, the model treated in this paper is generated as exactly one orbit on the parameter space. It is also shown that the BEE can be viewed as a generalized least squares estimator.