Spatial Scan Statistics for Multiple Clusters in Arbitrary Shapes

Spatial Scan Statistics for Multiple Clusters in Arbitrary Shapes

張志銘(Chih-Ming Chang)

指導教授 ： 林培生

國立中正大學/理學院/數學系統計科學研究所/碩士(2014年)

並列摘要

In this study, we propose a generalized scan statistic method with quasi-
likelihood function to simultaneously consider geographic clusters, covariates,
and spatial correlations for detecting multiple clusters. To improve the time-
consuming two-stage estimation process by Lin (2012), we first combine the
Kulldorff’s scan statistic method and variogram tool to estimate spatial cor-
relation, and then use the quasi-likelihood function to estimate coefficients
of geographic clusters and covariates. Instead of using the traditional likeli-
hood ratio test to detect cluster, we use the smallest p-value as a test statistic,
and apply resampling method to address the multiple testing problem. The
quasi-deviance criterion is used to regroup the estimated clusters for finding
arbitrary shapes of geographic cluster. For illustration, the method is applied
to enterovirus data from north Taiwan in 2003.Then we may discovery the
clusters of high disease area from the analysis.

並列關鍵字

Spatially correlated data ； Scan Statistics ； Quasi-likelihood estimate ； Multiple cluster ； Cluster detection

參考文獻

Chan, H. P. (2009), “Detection of spatial clustering with average likelihood

Cressie, N.A. (1993), Statistics for Spatial Data.

statistics for irregularly shaped clusters.”, Computational and Graphical

ratio test for spatial disease clustering.”, Statistics in Medicine, 20.

(2003), “A hierarchical model for spatially clustered disease rates.”,