摘要
应用EM算法求含缺失数据的约束线性模型回归系数的极大似然估计,该回归系数满足线性不等式约束.我们提出M-步的优化算法,并针对正态模型讨论EM序列的收敛性,最后举例说明算法的应用.
The paper gives the maximum likelihood estimators of regression coefficients in the constrained linear model which have missing data by using EM algorithm, here the constraints are given by linear inequalities. We propose an optimization algorithm of M-step and discuss some convergence properties of the EM sequence for the normal model. Finally, the suggested algorithm is illustrated by a numerical example.
出处
《南京大学学报(数学半年刊)》
CAS
2007年第1期122-131,共10页
Journal of Nanjing University(Mathematical Biquarterly)
关键词
缺失数据
EM算法
极大似然估计
线性不等式约束
missing data, EM algorithm, maximum likelihood estimators, linear inequalities constraints