不同空间权重矩阵在白血病空间统计分析中的探讨  

Study on the Spatial statistical analysis of leukemia using different spatial weight matrixes

作　　者：林冠文[1] 马金香[1] 

机构地区：[1]广州医科大学公共卫生学院统计学系,广东广州510182

出　　处：《中华疾病控制杂志》2013年第7期617-620,共4页Chinese Journal of Disease Control & Prevention

基　　金：广东省社会发展科技计划项目(2009B030801045)

摘　　要：目的探索不同空间权重矩阵在白血病发病的空间自相关性以及空间分布特征统计分析中的适应性及其应用价值。方法运用GeoDa0.9.5-i软件生成一阶、二阶、三阶Rook和Queen权重、距离阈值权重和K-Nearest权重矩阵,分别进行广东省儿童青少年白血病发病率的全局及局域空间自相关分析。结果 4种空间权重矩阵分析的全局Moran’s I值均有统计学意义。全局Moran’s I值在一阶Rook、Queen权重,最小距离阈值权重和K=3时分别为0.342、0.360、0.402和0.382(均有P<0.05),广东省儿童青少年白血病呈聚集性分布;4种空间权重矩阵在探测发病率低低聚集方面基本相同,在探测发病率高高聚集方面差别较大。结论结合不同的空间权重矩阵进行综合分析,有助于深入了解白血病发病的空间分布特征,为预防和控制儿童青少年白血病提供依据。Objective To analyze the spatial autocorrelation and to explore the spatial distribution pattern of leu- kemia using different spatial weight matrixes. Methods Based on first-order, second-order, third-order Rook and Queen spatial weight matrix,threshold weight matrix and K-nearest neighbor weight matrix, the spatial autocorrelation was carried out by GeoDa0. 9. 5-i. Results Four spatial weight matrix analysis of the global Moran＇s I values were statistically signifi- cant. The global Moran＇s I =0. 342, O. 360, O. 402 and 0. 382（ all P 〈0. 05） ,while in first-order Rook and Queen weight, smallest threshold and K = 3. Childhood leukemia cases was aggregated distribution in Guangdong Province. Four spatial weight matrixes in detecting low-low aggregation were basically the same, while detection of high-high aggregation varied greatly. Conclusions Combining different weight matrixes to make spatial autocorrelation analysis can explore the spatial distribution of leukemia and provide theoretical basis for further prevention and control.

关 键 词：白血病 数据分析 统计 流行病学研究 

分 类 号：R733.7[医药卫生—肿瘤] R181[医药卫生—临床医学]