金融时序数据建模中的模型设定问题  

Analysis of Econometric Modeling Selection in the Process of Financial Time Series Modeling

作　　者：黎实[1] 彭作祥[2] 庞皓[1] 

机构地区：[1]西南财经大学统计学院 [2]西南师范大学数学与财政学院

出　　处：《财经科学》2005年第3期61-68,共8页Finance & Economics

基　　金：教育部人文社会科学博士点基金项目资助 (批准号 :0 3JB790 0 11) ;国家自然科学基金项目资助(批准号 :70 3710 6 1) ;西南财经大学"十五""2 11工程"项目资助。

摘　　要：W .J.Granger与D .F .Hendry (2 0 0 4 )关于建模思路的对话引起了国际计量经济学界关于模型设定问题的争论,本文就这一问题分析讨论了在金融时序数据实证研究中得以广泛应用的ARCH/GARCH模型的设定问题,认为在金融时序数据的建模中,ARMA族模型不宜作为数据生成过程的模型设定,其统计性质也不能直接扩展到ARMA -GARCH族数据生成过程;虽然ARCH/GARCH族模型作为金融时序数据的生成过程有着良好的统计性质,但也不宜单纯采用一般到特殊的建模思路,而应是一般到特殊和特殊到一般两种建模思路的结合;ARCH/GARCH族模型的设定应当包含事前检验、事后检验等设定检验步骤。The dialog concerning econometric modeling between W. J. Granger and D. F. Hendry (2004) leads the controversy over model selection in the international econometric school. The paper focuses on the selection of ARCH/ GARCH family models in the context of financial time series data. The authors argue that traditional ARMA model family are not suitable to be proper model selection for financial data generating process, its statistic properties also not extended in process of model selection of ARCH/GARCH. Though ARCH/GARCH family with good statistical properties in modeling financial time series data, the general-to-simple approach is not good being appropriate, and the combination of the general-to-simple and simple-to-general approaches may be better choose. The procedure of ARCH/ GARCH model selection, including ante and ex post tests, is proposed finally.

关 键 词：计量经济学 模型设定 金融时序数据 ARMA族模型 ARCH族模型 GARCH族模型 

分 类 号：F224[经济管理—国民经济]