不同空间权重矩阵在疟疾空间分布模式分析中的探讨

目的探索不同空间权重矩阵在疟疾发病的空间自相关性以及空间分布模式中的适用性及其应用价值。方法采用Queen权重、距离阈值权重、K最邻近点权重生成不同的空间权重矩阵,应用GeoDa和R软件对疟疾的全局和局域空间自相关模式进行分析。结果 3种权重矩阵均显示2005-2007年云南疟疾全局Moran’sⅠ系数有统计学意义,2006年聚集性最强;局域显示3年均存在滇西部的＂正热点＂区域,2007年新增西北部＂正热点＂,部分＂负热点＂随着发病扩散逐渐有高发病趋势;全局Moran’sⅠ系数在最小距离阈值和K=3时取最大值,探测热点时距离阈值权重更易使聚集区域扩大,K最邻近点权重不容易发现低值聚集。结论应用不同的空间权重矩阵得到的热点区域存在一定差异,结合不同的空间权重矩阵进行分析,有助于加深对疟疾发病空间分布模式和流行蔓延趋势的认识。

An analysis of spatial distribution pattern of malaria using different spatial weight matrixes

Objective To analyze the spatial autocorrelation and to explore the spatial distribution pattern of malaria using different spatial weight matrixes.Methods Based on Queen weight matrix,threshold weight matrix and K-nearest neighbor weight matrix,the spatial autocorrelation was carried out by GeoDa0.9.5-i and R2.7.2 software.Results The three weight matrixes showed that the global Moran＇sⅠ coefficient was the largest in 2006,the statistic showed that there was a high-value cluster region in the west of Yunnan in the three continuous years,and a new high-value cluster region in the west-north,some low-value cluster regions had a trend of high incidence.The global Moran＇sⅠ coefficient had the largest value while the smallest threshold and K=3.The threshold weight tended to detect more cluster regions,and K-nearest neighbor weight was not good at detecting low-value cluster regions.Conclusions The cluster regions were different when using different weight matrixes.Combining different weight matrixes to do spatial autocorrelation analysis can explore the spatial distribution of malaria and provide theoretical basis for further prevention and control.