ARMA模型参数估计的格林函数法

论文根据一个适当高阶的AR模型可与一个ARMA模型等价的原理,推导了以格林函数为中间变量,建立ARMA模型的线性方法。用此方法对太阳黑子数据进行计算,所得的模型参数,与用非线性最小二乘方法的结果近似,且较用逆函数算出的参数精度高。实践证明这一方法在一定条件下足可行的。运用这一方法建立自回归滑动平均模型,可避免采用非线性最小二乘方法,只需进行很少几次的线性最小二乘便可建模,较传统方法,可大大减少运算工作盛,并便于在微机上推广应用。

PARAMETRIC ESTIMATION OF A RMA MODEL USING GREEN FUNCTION

According to the principle that an ARMA model is equivalent to a suitable AR model with higher order, an ARMA linear modeling mothod, using Green function as transfer variable, is presenled. As an example, Sunspot data have been calculated. The parameters of the model are si milar to those results obtained by the nonlinear leasL square methed, and have a higher accuracy than those results obtained by the inverse fune tion method, it has been proved Lhat this method is useful under some specific conditions. In comparison with classical methods without using the nonlinear least square method, the ARMA modeling can be done by only a few Limes of linear least square. The amount of calculation can be greatly reduced, and a microcomputer can be used.