基于多元单整变量之间协整性的Granger因果关系检验框架构建  

Construction of Granger Causality Test Framework Based on Co-integration of Multivariate Integral Variables

作　　者：靳庭良[1] Jin Tingliang(Henan University of Economics&Law,Zhengzhou 450002,China)

机构地区：[1]河南财经政法大学,郑州450002

出　　处：《统计与决策》2022年第11期50-55,共6页Statistics & Decision

摘　　要：文章通过将基于VAR模型的Granger因果关系检验问题等价地转化为基于差分模型(DM模型)或广义误差修正模型(GECM模型)中参数线性约束的F检验问题,建立了单整变量之间Granger因果关系的GECM-DM检验框架。研究发现:单整变量之间的(多项式)协整性(包括是否存在协整、协整的阶数和秩)会影响Granger因果关系检验依据的DM模型或GECM模型的设定;非Granger原因的原假设可能包含参数的线性约束式,其与变量之间的协整向量有关。对于一般的单整变量组,建立的规范实用的Granger因果关系GECM-DM检验框架,使Granger因果关系GECM-DM检验具有可操作性。This paper establishes the GECM-DM test framework of Granger causality between integral variables by equitably transforming Granger causality test based on VAR model into F test based on linear parameter constraints in difference model(DM model) or generalized error correction model(GECM model). Studies find that the(polynomial) co-integration between single integration variables(including the existence of co-integration, order and rank of co-integration) affects the setting of DM model or GECM model based on Granger causality test, and that the null hypothesis of non-Granger cause may contain linear constraints on parameters that are related to co-integration vectors between variables. For the general set of integral variables, a standard and practical Granger causality GECM-DM test framework is established, which makes Granger causality GECM-DM test operable.

关 键 词：GRANGER因果 协整 广义误差修正模型 原假设 

分 类 号：F224.0[经济管理—国民经济]