COMPSTAT 2018: Start Registration
View Submission - COMPSTAT2018
A0271
Title: Cluster detection for multi-dimensional spatial data based on hierarchical structure Authors:  Fumio Ishioka - Okayama University (Japan) [presenting]
Koji Kurihara - Okayama University (Japan)
Abstract: In recent years, the spatial scan statistics that detect a spatial cluster for spatial data is widely used in spatial epidemiology and other fields. However, currently, most of them are limited to applications on two-dimensional data such as geospatial data. Echelon analysis is an approach which enable us to visualize the spatial data systematically and objectively by a topological hierarchical structure according to the adjacency relationship of each region. Even if each region has multiple variables, if they are ordinal variables, and we can define relative positions between variables based on their order, it is possible to draw them as a two-dimensional dendrogram. Therefore, the echelon scan method that combines a spatial scan statistic and echelon analysis enables us to detect a spatial cluster for spatial data with multidimensional nature. We introduce how to perform the cluster detection for spatial data having multidimensional elements such as time series or multivariate by using the echelon scan method with showing a concrete example.