ARMA模型参数估计的两段最小二乘法

Two-stage Least Squares Method for Parameter Estimation of ARMA Models

提出了自回归滑动平均(ARMA)模型参数估计的两段最小二乘法。首先用递推最小二乘法对真实ARMA模型拟合高阶自回归(AR)模型,然后基于所拟合的AR模型参数,用最小二乘法解一个不相容代数方程组得到ARMA模型参数。一个仿真的例子说明了其有效性。

Two-stage least squares method for parameter estimation of ARMA models is presented. First, the true ARMA model is fitted by a high order AR model by using the recursive least squares method. Secondly, based on the fitted AR model parameters, the ARMA model parameter estimates are obtained via solving a set of contradictory algebriaic equations by using the least squares method. A simulation example shows its effectiveness.