二值数据下模型和异常点的同时识别  

A method of simultaneous model selection and outlier identification for binary data

作　　者：王康宁[1] 汪四水[1] 

机构地区：[1]苏州大学数学科学学院,江苏苏州215006

出　　处：《苏州大学学报（自然科学版）》2010年第4期9-13,共5页Journal of Soochow University(Natural Science Edition)

摘　　要：讨论了基于贝叶斯方法进行模型选择与异常点识别时两者之间的相互影响,建议模型与异常点应结合起来同时识别.针对二值数据,采用引入隐变量的数据扩增方法进行异常点识别,并且给出了基于MCMC方法计算后验概率来进行模型和异常点同时识别的具体过程.In this paper we discuss the interaction between model selection and outlier identification based on bayesian method.We suggest that model selection and outlier identification should be identified simultaneously.For binary data,data augmentation method is introduced to identify the outlier,and we suggest a method for simultaneous model selection and outlier identification based on posterior probabilities computed by MCMC method.

关 键 词：二值数据 隐变量 模型选择 异常点识别 GIBBS抽样 

分 类 号：O212[理学—概率论与数理统计]