This paper puts forward a long memory stochastic conditional durations ( LMSCD) model for ultra-high frequency ( UHF) durations series, and provides a kind of spectrum likelihood estimation method. 文章针对股票市场的超高频持续期序列,提出了长记忆随机条件持续期模型(LMSCD),并给出了模型参数的极大谱似然函数估计方法。
The state space Kalman filter recursive computation and GARCH model's conditional variance recursive computation are used to estimate the conditional likelihood of parameters. Akaike's minimum AIC procedure is used to select the best model fitted to the data within and between the alternative model classes. 参数的条件最大似然估计应用了状态空间模型的卡尔曼滤子递推和GARCH模型的条件方差递推,模型阶数的选取应用了Akaike的最小化信息矩阵方法。