泊松INAR(1)模型参数估计方法的比较及其应用

Comparison of Parametric Estimation Methods for Poisson INAR(1) Model with Its Applications

探讨了泊松INAR(1)模型的参数估计.基于CLS估计方法,提出一种偏差修正CLS估计方法并分析了其大样本性质.随机模拟结果表明:与CLS估计和CML估计方法相比,偏差修正CLS估计方法可以有效减小参数估计的偏差.偏差修正CLS估计方法在有限样本下估计效果较好,同时计算过程简便,具有较高的实用性.最后,通过泊松INAR(1)模型对一组实际的整数值时间序列数据建模,利用偏差修正CLS估计方法进行参数估计,并对模型进行预测.

The parametric estimation methods are considered for the INAR（1） model with Poisson distributed innovations. Based on the CLS estimation method, the bias-adjusted CLS estimator for the Poisson INAR（1） model is derived. The asymptotic properties of the bias-adjusted CLS estimator are also studied. A series of stochastic simulations show that the proposed estimation method has the best performance in terms of bias, comparing with CLS estimation method and CML estimation method. The bias-adjusted estimator can be conveniently used in empirical studies due to its good finite-sample behavior and simple computation procedure. Finally, the bias-adjusted CLS estimation method to a real time series data is applied, which can produce preferable forecasts.