指数分布异常数据检验的Bayes方法  

Test of outliers by Bayesian

作　　者：肖浩[1] 刘次华[1] 董锐[1] 黄丽琨[1] 

机构地区：[1]华中科技大学数学系,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2004年第8期92-94,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目 (70 2 71 0 6 9)

摘　　要：研究了指数分布场合异常数据的检验问题 ,提出了无信息先验分布下的Bayes检验方法 .利用分布参数的先验分布严格推导出新的检验统计量及其后验分布函数 .实例和蒙特卡罗模拟表明了方法是可行的 ,适用于同时含有异常大、异常小数据的情形 。The paper studied the traditional tests of an exponential sample and presented a Bayesian test based on the noninformative prior distribution. By using prior distribution of the distributional parameter, test statistics and their post distributional functions were derived from the strict deduction. The example and Monte carlo simulation experiments showed its effectiveness. The test was suitable for the data including upper and lower outliers and could avoid the masking effect of the data.

关 键 词：异常数据 指数分布 BAYES方法 先验分布 数据检验 

分 类 号：O212.8[理学—概率论与数理统计]