自回归模型参数的结构总体最小二乘估计

摘要

结构总体最小二乘法不仅考虑系数矩阵误差,而且顾及系数矩阵(或与观测矢量组成的增广矩阵)中重复元素必须具有相同改正数的约束,平差精度比最小二乘法高,理论比总体最小二乘法更加严密.采用间接平差理论推导在观测矢量和系数矩阵都含有相同重复元素情况下的结构总体最小二乘迭代算法；最后通过数字实验证明该算法的可行性,有效性以及该方法实现自回归模型参数估计的优越性和可靠性.%The structured total least squares not only take into account the errors of coefficient but also consider the identical elements have attained the same correction value in coefficient or in argument of vector of observation L and coefficient matrix A. This method make the adjustment theory more strict than others such as least squares and total least squares. In this contribution that we give an simply iteration algorithm for structured total least squares, and give a numerical example to demonstrate the efficiency and feasibility, as well as the advantages of the structured total least squares for AutoRegresion parameters estimation at last .