摘要
为了诊断具有异方差的线性回归模型的异常点,建立了具有异方差的均值漂移模型和数据删除模型.采用Score诊断统计量对具有异方差的均值漂移模型的均值是否漂移进行诊断,证明了异方差存在条件下均值漂移模型和数据删除模型的等价性.这一结果表明,在诊断具有异方差的线性回归模型的异常点时,可考虑采用更加便于处理的均值漂移模型.最后,用Score诊断统计量对镀锌数据进行了异常点的诊断.
The mean shift outlier model （MSOM） and the case deletion model （CDM） with heteroscedasticity were developed in order to detect outliers in linear regression models with heteroscedasticity. The diagnostic test for the mean of the MSOM with heteroscedasticity was carried out based on the score statistic. It was proven that the estimates of MSOM and CDM with heteroscedasticity are equal. This result shows that the MSOM, which is easier to process, can be used more often to detect outliers in the linear regression model with heteroscedasticity than CDM. Finally, outliers in galvanization data were detected using the score statistic.
出处
《河海大学学报（自然科学版）》
CAS
CSCD
北大核心
2008年第2期279-282,共4页
Journal of Hohai University(Natural Sciences)
关键词
均值漂移模型
数据删除模型
异方差
mean-shift outlier model
case deletion model
heteroscedasticity
