Premium
A Spatial‐EWMA Framework for Detecting Clustering
Author(s) -
Lin Chenju,
Tsui KwokLeung,
Lin Chenyu
Publication year - 2014
Publication title -
quality and reliability engineering international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.913
H-Index - 62
eISSN - 1099-1638
pISSN - 0748-8017
DOI - 10.1002/qre.1484
Subject(s) - ewma chart , data mining , multivariate statistics , spatial analysis , cluster analysis , computer science , statistics , mathematics , artificial intelligence , process (computing) , control chart , operating system
Spatial surveillance is critical to health systems, manufacturing industries, and in many other domains. For example, determining hotspots of infectious diseases and detecting defect patterns on semiconductor wafers require sensitive spatial analysis tools. The goal of this paper is to detect spatial clusters with mean shifts. Conventional multivariate analysis methods may ignore spatial structure among data and lead to inefficient inspection. Several likelihood ratio‐based scan statistics have been designed for spatial surveillance. However, there is no most powerful test when parameters like shift magnitude and coverage are unknown. This paper proposes a spatial exponentially weighted moving average (spatial‐EWMA) approach that can detect the existence and locate the potential centers of shift clusters. The test procedure assigns different weights to the data with different radius levels from the investigated shift center. The efficiency of the spatial‐EWMA approach is shown by simulation. Lastly, an example of detecting the counties with high incidence of male thyroid cancer in New Mexico is provided to show the effectiveness of the proposed approach. Copyright © 2013 John Wiley & Sons, Ltd.