首页 | 本学科首页   官方微博 | 高级检索  
     检索      

有限混合Laplace分布回归模型局部估计的EM算法
引用本文:王继霞,汪春峰,苗雨.有限混合Laplace分布回归模型局部估计的EM算法[J].数学杂志,2016,36(4):667-675.
作者姓名:王继霞  汪春峰  苗雨
作者单位:河南师范大学数学与信息科学学院, 河南 新乡 453007,河南师范大学数学与信息科学学院, 河南 新乡 453007,河南师范大学数学与信息科学学院, 河南 新乡 453007
基金项目:Supported by National Natural Science Foundation of China (11471104); Nat-ural Science Foundation of Henan Educational Committee (2011B110018); Program for Innovative Re-search Team (in Science and Technology) in University of Henan Province (14IRTSTHN023).
摘    要:本文研究了一类有限混合Laplace分布回归模型的局部极大似然估计问题. 利用核回归方法和最大化局部加权似然函数的EM算法, 获得了参数函数的局部极大似然估计量, 并讨论了它们的渐近偏差, 渐近方差和渐近正态性. 推广了有限混合回归模型下局部非参数估计的结果.

关 键 词:有限混合模型  Laplace分布  EM算法  局部极大似然估计  核回归
收稿时间:2013/12/8 0:00:00
修稿时间:2015/2/26 0:00:00

THE EM ALGORITHM OF LOCAL ESTIMATIONS FOR FINITE MIXTURE OF REGRESSION MODELS WITH LAPLACE DISTRIBUTION
WANG Ji-xi,WANG Chun-feng and MIAO Yu.THE EM ALGORITHM OF LOCAL ESTIMATIONS FOR FINITE MIXTURE OF REGRESSION MODELS WITH LAPLACE DISTRIBUTION[J].Journal of Mathematics,2016,36(4):667-675.
Authors:WANG Ji-xi  WANG Chun-feng and MIAO Yu
Institution:School of Mathematics and Information Sciences, Henan Normal University, Xinxiang 453007, China,School of Mathematics and Information Sciences, Henan Normal University, Xinxiang 453007, China and School of Mathematics and Information Sciences, Henan Normal University, Xinxiang 453007, China
Abstract:In this paper, we study the local maximum likelihood estimations for a dinite mix-ture of regression models with Laplace distribution. By using the kernel regression method and the EM algorithm for maximizing the local weighted likelihood functions, we obtain the local maximum likelihood estimations of the parametric functions, and discuss their asymptotic biases, asymptotic variances and asymptotic normality, which extend the results of the local non-parametric estima-tions for the finite mixture of regression models.
Keywords:finite mixture model  Laplace distribution  EM algorithm  local maximum like-lihood estimation  kernel regression
点击此处可从《数学杂志》浏览原始摘要信息
点击此处可从《数学杂志》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号