SELECTING ORDER FOR GENERAL AUTOREGRESSIVE MODELS BY MINIMUM DESCRIPTION LENGTH
针对无特征多项式根约束的一般自回归模型，基于最小描述长度提出阶数估计准则，并证明估计的强相合性。
Abstract. In this paper we put forward some criteria for estimating the order for general autoregressive (AR) models (i.e. AR models without any constraint about the roots of the characteristic polynomial) according to the minimum description length. Different criteria are given for different kinds of AR models because the convergence rates are different. It is proved that all the estimates for the order are strongly consistent.