带趋势项GARCH误差模型参数变化的残量检验  

Residual CUSUM Tests for Parameters' Change in GARCH Error Models with Deterministic Trend

作　　者：金浩[1] 田铮[1,2] 张思[3] 

机构地区：[1]西北工业大学应用数学系,西安710072 [2]遥感科学国家重点实验室,中科院自动化所,北京100101 [3]西安科技大学高新学院,西安710109

出　　处：《工程数学学报》2009年第6期1069-1082,共14页Chinese Journal of Engineering Mathematics

基　　金：国家自然科学基金(50905140;60375003);西北工业大学科技创新基金(2007KJ01033)

摘　　要：本文研究带趋势项GARCH误差模型的参数变点检验。在参数未知的情况下,利用残量累积和检验以消除相依序列对检验结果的影响:研究表明在原假设条件成立下,残量累积和统计量的极限分布是一个标准布朗桥的上确界。蒙特卡罗数值模拟和IBM收益率数据分析的结果都充分说明了本文方法的有效性和实用性。This paper analyzes the test problem for parameters＇ change in GARCH error models with deterministic trend. The residual cusum test is proposed when the parameters are unknown, which can weaken the influence of dependent observations, It is shown that the asymptotically limiting distribution of the residual cusum test statistic is the supermum of a standard Brownian bridge under the null hypothesis. In order to verify our conclusion, we carry out a Monte Carlo simulation and examine the return of IBM data. The results from both simulation and real data analysis support our claim.

关 键 词：残量累积和检验 不变性原理 布朗桥 变点 

分 类 号：O212.9[理学—概率论与数理统计]