具有ARIMA(0，1，0)对称误差的非线性模型的统计诊断  

Statistical Diagnostics in Nonlinear Models with ARIMA(0,1,0) Symmetrical Errors

作　　者：曹春正[1] 林金官[1] 朱晓欣[2] 

机构地区：[1]东南大学数学系,南京210096 [2]南京信息工程大学数理学院,南京210044

出　　处：《应用数学》2008年第3期490-497,共8页Mathematica Applicata

基　　金：国家自然科学基金(10671032);南京信息工程大学校科研基金

摘　　要：本文讨论具有ARIMA(0,1,0)对称误差的非线性模型的异方差检验和局部影响分析.对称误差分布族包括正态,t,power exponential,logisticsⅠ,Ⅱ,污染正态等所有对称连续分布.文章首先导出了关于白噪声异方差检验的score统计量及其调整形式,然后对模型进行了局部影响分析,得到了基于似然函数扰动和反应变量扰动的诊断统计量.最后,利用实际数据说明了检验方法的应用,并用Monte Carlo模拟方法研究了异方差检验统计量的检验功效.In this paper,we discuss the heteroscedastic test and local influence in nonlinear models with ARIMA （0, 1, 0） symmetrical errors. The symmetrical errors distribution class includes all symmetrical continuous distributions such as normal, Student-t, power exponential,Logistics Ⅰ and Ⅱ ,contaminated normal, so on. First, score test statistic and its adjustment form of the heteroscedastic test are derived. Then,we discuss the local influence analysis of the model and derive diagnostic statistics which based on likelihood functions perturbation and response perturbation. Finally, a real data is used to illustrate the applying of our testing methods, and the properties of test statistics are investigated through Monte Carlo simulations.

关 键 词：对称分布 ARIMA(0 1 0) 异方差 SCORE检验 局部影响分析 

分 类 号：O212.2[理学—概率论与数理统计]