乙脑空间分布异质性和相关性的多水平MMMC模型分析

Analysis of Spatial Heterogeneity and Autocorrelation of Geographical Distribution of Japanese Encephalitis by Applying Multilevel MMMC Model

目的乙脑空间分布存在地区差异和聚集性,传统统计模型处理这类数据存在缺陷,探讨多水平MMMC模型应用的可行性,为复杂空间数据建模提供方法学参考。方法对2004～2007年31个省(自治区、直辖市)乙脑分布数据分别采用传统统计模型、多水平模型、多水平MMMC模型分析,比较模型拟合结果。结果多水平MMMC模型的拟合效果最好,分离的区域效应和近邻效应均有统计学意义(P<0.05),水稻种植面积比例、生猪出栏数、农村人口比例对乙脑地区分布有影响(P<0.05)。结论多水平MMMC模型可以分离和比较疾病空间分布的区域效应和近邻效应,探索疾病分布地区差异和聚集性原因。

Objective It is reported that spatial heterogeneity and correlation exist in geographical distribution of Japanese Encephalitis（JE） occurring.Traditional regression models have obvious weakness on dealing with those spatial data.Multilevel MMMC model is introduced as an appropriate way of handling data on such complex structure in purpose of methodological reference.Methods Traditional Poisson regression model,multilevel Poisson model and multilevel MMMC model are built and compared to analysis the data of JE’s geographical distribution of 31 provinces （autonomous region or borough） from 2004 to 2007.Results Multilevel MMMC model has best goodness of fit in those models.Separately random effect of areas and a multiple membership effect term for surrounding neighbour areas have statistical significance（P0.05）,and proportion of rice planting,pig rearing,proportion of peasant population have effect on geographical distribution of JE（P0.05）.Conclusion Multilevel MMMC model can estimate random effect of areas and neighbour areas in spatial distribution of disease simultaneously,and reveal possible reasons how to induce spatial heterogeneity and cluster.