ARMA模型的几种定阶方法.pdfVIP

摘 要 本文针对自回归滑动平均（ARMA ）过程的参数估计和定阶，把 Box-Jenkins 方法 和 BIC 准则与系统的定阶方法加以结合，详细地介绍了基于线性估计和自回归定阶准则 的方法(ARCRI)和公因子检验定阶改进法 。并且把这两种方法和广义样本自相关系数函 数定阶法（ESACF ）加以比较总结。 ARCRI 在估计部分采用了两步回归法（two-stage regression method ），用自回归定 阶准则来确定模型的阶。 而公因子检验定阶改进法的参数估计是基于长自回归系数所组成的矩阵，模型阶数 的确定则着眼于上述矩阵行列式是否为零。 在方法的比较中，主要从样本大小，模型结构和参数值大小对模型合理性的影响三 方面加以阐述。 关键词：Box-Jenkins 方法；BIC 准则线性估计；自回归定阶准则；公因子检验 I Abstract This article aims at the estimation and order identification of the autoregressive moving average (ARMA) models. As performing Box -Jenkins method BIC and system order identification to unify we introduced the Autoregressive order determination criterion and common factor test in detail. And compare these two methods to ESACF at the end. The former used two-stage regression method in estimation and the Autoregressive order determination criterion in order identification. In the latter , however ,both estimation and identification are based on a matrix formed from the coefficients of autoregressive approximation to the process of interest. We show that a zero determination of this matrix is necessary and sufficient for the existence of a common factor in autoregressive and moving average lag polynomials and therefore for redundant parameters in the model. In the method comparison, mainly the influence performs from the sample size , model structure and parameter values to the rationality of the model is elaborated Key words : Box -Jenkins method ; Bayesian information criterion ; two-stage regression method ; Autoregressive order determination criterion ; common factor test II