估计SURE模型参数的一种方法探讨

PROBE INTO AN APPROACH OF ESTIMATING PARAMETERS IN SURE MODELS

提出了计算ＳＵＲＥ（ＳｅｅｍｉｎｇｌｙＵｎｒｅｌａｔｅｄＲｅｇｒｅｓｓｉｏｎＥｑｕａｔｉｏｎｓ—表面无关回归方程）模型参数的一种方法。利用正交变换，以避免在直接计算方差－协方差逆矩阵（或其估计）时，经常导致的不必要的精度损失的困难。较详细地考虑了特殊的ＳＵＲＥ模型的求解过程，在这里，每个方程的回归变量集中包含前面方程的回归变量作为当前回归变量集的适当子集。

An approach to compute the parameters of a seemingly unrelated regression equation (SURE) models is proposed. Orthogonal transformations are employed to avoid the difficulties in directly computing the inverse of the variance covariance matrix (or its estimate) which often lead to unnecessary loss of accuracy. The procedure for solution of the special SURE models, in which the problem is constrained so that the regressors in each equation contain the regressors in previous equations as a proper subset, is considered in detail.