Abstract

The purpose of this article is the statistical surveillance of spatial autoregressive models, where the observed process is monitored over both space and time. The considered spatial model contains disturbances with heavy tails. The control procedures based on exponential smoothing or cumulative sums are constructed using characteristic quantities including the first and the second moments to monitor both means and covariances. Via Monte Carlo simulation, the in-control upper control limits of the control schemes are derived. In a further simulation study, we compare the detection speed of these procedures in the out-of-control situation.

中文翻译：

---

重尾空间自回归模型统计监测的时空程序

摘要

本文的目的是对空间自回归模型进行统计监控，其中观察到的过程在空间和时间上都受到监控。所考虑的空间模型包含重尾扰动。基于指数平滑或累积和的控制程序是使用包括一阶和二阶矩的特征量来构建的，以监控均值和协方差。通过蒙特卡罗模拟，导出了控制方案的在控上限。在进一步的模拟研究中，我们比较了这些程序在失控情况下的检测速度。