AR(p)模型参数估计方法比较和实证分析

On comparison and empirical analysis of AR(p) model parameter estimation Methods

对时间序列AR(p)模型的参数估计YULE-WALKER法、最小二乘法、最大似然法三种估计方法进行分析和比较,从模型推导出最小二乘估计、最大似然估计实质上是同一种估计方法,应用MATLAB软件对我国CPI数据建立AR(2)模型并应用3种方法对其参数进行估计、模型检验、预测结果比较,得出的结论与理论推导相符,实证上说明AR(p)模型参数估计使用最小二乘法和最大似然法估计的结果是一样的。

By analyzing and comparing YULE WALKER, least squares method and maximum likelihood estimation method of time series AR （p） model, least squares method and maximum likelihood estimation method are found essentially the same kind of estimate method. Building China＇s CPI data AR （2） model by using MATLAB software and then applying three methods to estimate the parameters, check the models and compare the prediction results,the conclusion obtained are found consistent with the theoretical deriva lion. The empirical analysis shows the estimated results of AR （p） model parameter are the same by using least squares method and maximum likelihood method.