摘要
对于一般的生长曲线模型Y=ABC+ε,其中Vec(ε)具有0期望向量和协方差矩阵V W,当设计矩阵具有复共线性关系时考虑了参数矩阵B的岭估计问题。根据实际背景假设误差阵的协方差阵可未知,此时用传统的典则化方法很难处理在参数估计过程中出现的问题。采用奇异值分解,得出了广义岭估计类F的相关性质,并用极小化均方误差的无偏估计法、Hemmerle-Brantle估计法,以及Q(c)准则讨论了岭参数阵K的选取。
For the growth curve model with mean vector and dispersion matrix , the problem of estimating parameters matrix is considered. In general, dispersion matrices of error - matrix may be unknown in practice. In these cases it is difficult to solve some of the raised problems. In the paper, the singular values decomposition method is used to obtain some further conclusions on generalized ridge estimates class. In addition, some methods of choosing matrix are given, including minimizing unbiased estimate of the mean squared error and Q （ c ） criterion.
出处
《淮阴工学院学报》
CAS
2008年第3期73-79,共7页
Journal of Huaiyin Institute of Technology
关键词
生长曲线模型
广义岭估计
奇异值分解
Q(c)准则
growth curve model
generalized ridge estimate
singular value decomposition method
Q （ c ） criterion
