A0422
Title: A novel method for spatial cluster detection in continuous data
Authors: Fumio Ishioka - Okayama University (Japan) [presenting]
Yusuke Takemura - Kyoto Women's University (Japan)
Koji Kurihara - Kyoto Women's University (Japan)
Abstract: Spatial phenomena concentrated in specific regions, such as the mortality rate of certain diseases across municipalities, are referred to as "clusters". In recent years, in fields like spatial epidemiology, spatial scan statistics have been widely employed to explore specific regions and evaluate the presence of clusters using likelihood-based methods. This test comprises two elements: 1) a statistical model and 2) a scanning method, which are combined for analysis. However, current spatial scan statistics predominantly employ Poisson models for counting data (discrete values), such as the number of disease cases or traffic accidents, on the statistical model front. Alternatively, while weighted Normal models accommodating regional variations are suggested for continuous data, circular scan methods remain predominant. These methods mainly focus on concentric scanning of regions, presenting difficulties in accurately identifying non-circular clusters, like those following rivers or roads. Therefore, to detect clusters of arbitrary shapes when analyzing continuous values, it is endeavored to apply the echelon scan method developed to the scanning approach. Furthermore, the aim is to validate how the application of this method influences the accuracy of cluster detection.
