SCD模型与ACD模型比较研究

Comparative Study of SCD and ACD Sodels

针对近几年在研究金融市场超高频序列时出现的ACD模型和SCD模型,先从理论上探讨了ACD模型、SCD模型与ARMA模型之间的关系,指出两类模型均可转化为ARMA模型,具有一定的相通性;然后实证比较了两类模型的自相关函数对实际数据自相关系数的刻画能力,以及利用基于随机模拟的似然比检验方法,从实证角度比较两类模型对持续期序列的拟合优度,得出在拟合金融市场超高频持续期数据时,SCD模型比ACD模型更具有优势。

Relationship among ACD and SCD models, which appeared in financial market＇s ultra high frequency data in recent years, and ACD model were explored theoritically. The results indicate that these two kinds of the models can be transformed to ARMA models, there being some common characters between them. For the two kinds of the models, their depicting abilities for autocorrelation functions were compared empirically and their good of fit were compared by the likelihood ratio test based on stochastic simulation. It is concluded that SCD model has more advantage than ACD models.