摘要以GPS测站坐标残差序列为实验对象,采用ARMA模型两种经典建模方法(Box-Jenkins方法和Pandit—Wu方法)进行了建模拟合的比较与分析。实验比较发现,当残差数据比较丰富时两种建模方法拟合效果理想,无明显差异,当实验数据较少时,Pandit—Wu方法拟合效果更好一些,剩余残差振幅较Box—Jenkins方法减少约1/3,且分布更加平稳。
As to GPS station coordinate residuals, we used two classical ARMA modeling methods (Box-Jen- kins method and Pandit-Wu method) to fit the GPS station coordinate residuals, and made comparative study. According to the experiment, we found out that the two methods have a little difference while there is abundant data. If the data is poor, with the Pandit-Wu method the results will be better. The amplitude of residuals reduces about 1/3 and the distribution is more uniform.