含有重度缺失的多维时间序列补缺方法及其在环境监测中的应用

Amending Method for Multi-Dimensional Time Series with Heavy Missing Data and Its Application in Environment Monitoring

目的 探讨解决环境污染监测资料中含有重度缺失的多维时间序列的补缺问题。方法 构造带有ARMA误差的线性方程组模型对缺失值进行迭代估计 ,通过基于带有输入序列的AR(1)模型的模拟研究和基于CO等环境监测数据的应用研究 ,与其他常用的 3种补缺方法 ,即多元线性回归、单个ARIMA模型和 3次样条插值法相比较 ,分析各方法的优劣。有关计算用SAS统计软件编程实现。结果 带有ARMA误差的线性方程组模型方法明显优于其他3种方法。结论 本研究提供的方法适合解决含有重度缺失的多维时间序列补缺问题 ,可以在环境监测领域应用。

Objectives To explore a better amending method for a set of correlated time series with heavy missing data and its application to monitoring data on environment pollution.Methods A model of Simultaneous Linear Equations with ARMA Error is proposed for iterative estimation of missing values.Simulation study is performed on the basis of an AR(1)model with input series and applied study is performed on th basis of a real montoring data set on CO-pollution respectively.The results are compared with those estimated wih three usually used apporaches,Multivariate Linear Regression,Simple ARIMA and interpolation by cubic spline in terms of statistical property.Programming is performed with statistical software SAS.Results The iterative approach based on the model of Simultaneous Linear Equations with ARMA Error is better than other three approaches significantly.Conclusion The amending method proposed can be used to estimate the missed values of a set of multi-dimensional time series with heavy missing,and apply to relevant monitoring data in real life.