ROBUST ESTIMATION OF A NORMAL MIXTURE

作     者：DEVEAUX, RD KRIEGER, AM 

作者机构：UNIV PENNWHARTON SCHDEPT STATPHILADELPHIAPA 19104 

出 版 物：《STATISTICS & PROBABILITY LETTERS》 (统计学与概率论通讯)

年 卷 期：1990年第10卷第1期

页      面：1-7页

核心收录：

学科分类：07[理学] 0714[理学-统计学（可授理学、经济学学位）] 0701[理学-数学] 070101[理学-基础数学] 

主　　题：Iteratively reweighted median iteratively reweighted MAD EM algorithm maximum likelihood estimation 

摘      要：A robust procedure for estimating the five parameters of a mixture of two normals is proposed. The procedure is based on a robustification of the steps of the EM algorithm. This results in the use of an iteratively reweighted median and MAD. Simulation results are presented that show the superiority of the robustified EM when compared to the usual EM algorithm estimates.