遗传参数的极大似然估计  

Maximum Likelihood Estimation of Genetic Parameters

作　　者：张晴霞[1] 刘妍岩[2] 

机构地区：[1]西南石油大学理学院,四川南充637001 [2]武汉大学数学与统计学院,湖北武汉430072

出　　处：《武汉大学学报（理学版）》2009年第6期646-650,共5页Journal of Wuhan University:Natural Science Edition

基　　金：国家自然科学基金资助项目(10771163)

摘　　要：将Luo和Woolliams模型中的混合正态分布推广到混合对数正态分布,先用矩量法得到各参数的估计值,再将此估计值作为EM(Expectation Maximization)算法迭代的初值进行迭代,将矩量估计和极大似然估计结合起来,提高了EM算法的收敛速度并使修改后的模型更符合生物学事实.In QTL （quantitative trait locus） studies, it is commonly assumed that the quantitative trait are distributed as normals. However, as Luo ＆ Woolliams have pointed, this assumption is not reasonable in practice, which will lead to bias of the corresponding estimators. They also have suggested that the log- normal distribution of quantitative trait would be more reasonable. In this paper, under the assumption of lognormal distributed quantitative trait, we follow two estimating methodologies proposed by Luo and Woolliams： moment estimate and maximum likelihood estimate based on EM algorithm to estimate parameters.

关 键 词：极大似然估计 EM(Expectation Maximization)算法 对数正态 数量性状基因位点 连锁 标记基因 

分 类 号：O212[理学—概率论与数理统计]