生长曲线模型中回归系数的新根方估计  

A New Root Power Estimator of Regression Coefficients in a Growth Curve Model

作　　者：方芳[1] 江舜君[2] 

机构地区：[1]南京工程学院基础部,江苏南京211167 [2]南京森林公安高等专科学校信息技术系,江苏南京210046

出　　处：《南京工程学院学报（自然科学版）》2007年第3期8-15,共8页Journal of Nanjing Institute of Technology(Natural Science Edition)

摘　　要：针对生长曲线模型中设计阵A与C至少有一个病态时的情况提出新根方估计■(m1,m2),并证明通过新根方参数mi(i=1,2)的适当选取,可使得该估计在均方误差意义下优于最小二乘估计及普通根方估计,证明其容许性及其对最小二乘估计抗干扰性的改进,并给出确定参数mi的两种方法.A new root power estimator B（m1, m2） is proposed as to solving the problem that either or both of design matrixes A and B of growth curve model is/are likely to be ill-conditioned. It proves that an appropriate selection of new root parameters ensures that this estimator is superior to LS estimator and normal root estimator when taking mean square errors into consideration. Furthermore, the improvement with regard to the admissibility and anti-interference to LS estimator was confirmed. Two methods of selecting new root power parameters mi were introduced.

关 键 词：生长曲线模型 LS估计 根方估计 均方误差 可容许性 

分 类 号：O212[理学—概率论与数理统计]