局部聚类的鉴别：一种基于模型的Moran's I检验  

Identification of local clusters:a model-based Moran's I test

作　　者：张海森[1] 宋向东[1] 于尚洋[1] 杨婧[1] 

机构地区：[1]燕山大学理学院,河北秦皇岛066004

出　　处：《燕山大学学报》2007年第6期545-548,557,共5页Journal of Yanshan University

摘　　要：传统意义上的Moran's I只能检验出是否存在聚类,但不能检验出是否存在局部聚类。为了弥补这项不足,在对数线性模型的基础上作Moran's(?)残差检验,记作I_(dr)。在非齐性总体下,I_(dr)比传统(?)有较好的第一类错误概率。然后,在原模型中加入空间关联项重做I_(dr)检验。当原模型中I_(dr)是显著的,但加入空间关联项后I_(dr)不显著,表明仅存在局部聚类:当原模型中I_(dr)是显著的,但加入空间关联项后I_(dr)仍显著,表明存在全局聚类。Traditional Moran＇s I can identify the cluster trend, but it can＇t identify the local clusters trend. In order to make some advances, a loglinear model-based test statistic for Poisson count data is set out, that is Idr. Idr is effective in reducing type I error probabilitiesofthe traditional Moran＇sldue to heterogeneous population size. When a significantlIr is contributed mainly by a local cluster, a spatial association term to remove the cluster effect can be devised. Since Idr is sensitive to the existence of local clusters but not sensitive to the presence of the global trend, the inclusion of a spatial association term in the Idr test can indicate if a global clustering trend exists or not.

关 键 词：对数线性模型 非齐性总体:空间关联项:局部聚类 

分 类 号：O212.7[理学—概率论与数理统计]