Mixture periodic autoregressive time series models

Mixture periodic autoregressive models are introduced to fit periodic time series with asymmetric or multimodal distributions. The stationary conditions of such series are derived, the asymptotic property of maximum likelihood estimators is obtained, and the application of EM algorithm is discussed. The new model class is illustrated by analyzing the particulate matter concentrations in Cleveland, OH.Periodically correlated time series Periodic autocovariances Mixture periodic autoregressive models EM algorithm