Abstract: In combination with finite mixture modeling, mixture periodical autoregressive moving average (MPARMA) models were introduced to fit periodic time series with asymmetric and multimodal distributions, the stationary condition of such series was derived, and the application of Expectation Maximization (EM) algorithm was discussed. The new model was evaluated by analyzing the PM10 concentrations.

Key words: periodic correlated time series, Periodical Autoregressive Moving-Average (PARMA), stationary, Expectation Maximization (EM) algorithm, heteroskedasticity

摘要： 为了描述周期时间序列中的偏倚和多峰等非线性特征，结合有限混合模型方法，提出混合周期自回归滑动平均时间序列模型（MPARMA），给出了MPARMA模型的平稳性条件，讨论了期望最大化（EM）算法的应用，通过PM10浓度序列分析，评估了MPARMA模型的表现。

关键词: 周期时间序列, 周期自回归滑动平均, 平稳性, EM算法, 条件异方差