利用SAS中非线性混和效应模型分析城市中阿拉伯按蚊水体滋生环境的时空变异系数

Uniform Convergence of Ergodic Markov Chains Using Gaussian Quadratures in SAS PROC NLMIXED for Calculating Marginal Likelihoods in Space Time-Varying Coefficients of Urban Anopheles gambiae S. L. Aquatic Habitats

目前对于时间序列分析中出现的空间自相关现象，已经有很多的的文献详述了如何对这种现象进行识别与建模分析。但是，对于撒哈拉以南非洲地区疟疾的主要媒介-阿拉伯按蚊幼虫滋生地的地理空间相关性方面的研究还没研究报道。在空间回归分析中，空间相关性主要通过空间滤波方法完成，该滤波方法通过分析时间序列中随机误差的分布特征。运用基于地理连接矩阵的非参数特征向量分解空间滤波方法，以寻求能鉴别阿拉伯按蚊幼虫滋生水生环境的预测因子。首先，选择肯尼亚的两个城市，Kisumu和Malindi，进行了蚊虫滋生水体的现场采样工作。然后获取快鸟（QuickBird）卫星数据；并利用该数据对冈比亚按蚊的滋生水体进行遥感影像分析。把这些水体的地理空间信息数据输入SAS／GIS模型中，进行单因素分析、相关分析、数据分布分析、以及全局空间自相关统计。然后，利用空间协方差参数生成局部空间自相关指数（如：Moran’s指数）。并由该模型生成了选择性特征向量的粗子集，此粗子集可以被用来研究这些样方变量间的空间互作效应。最后，利用贝叶斯系数估计法定义矩阵中各参数的期望先验参数值，得知水体环境的深度是决定该水体能否滋生阿拉伯按蚊的关键预测性因素，并且深度与其滋生可能性呈明显的正相关关系；由转移空间连接矩阵生成的特征向量的特殊子集可以获知由阿拉伯按蚊水生环境预测因子以及所建立的空间回归模型中空间误差的相关性。

Many texts on time series detail the way temporal autocorrelation can be identified and modeled. Literatures on spatial autocorrelation for georeferenced aquatic habitats of Anopheles gambiae s. l. , a major vector of malaria in Sub- Saharan Africa （SSA）, are less well developed. One approach to dealing with spatial autocorrelation in regression analysis involves filtering, which seeks to undertake data analyses within the context of the study of stochastic error characteristic of time series analysis. This article uses a nonparametric eigenfunction decomposition spatial filtering approach based on the geographic connectivity matrix for identifying predictors associated with productive An. gambiae s. 1. aquatic habitats. Initially, field sampling of aquatic habitats were performed in two urban environments, Kisumu and Malindi Kenya. QuickBird visible and near-infrared data at 0. 61 m spatial resolution data was acquired and was used to synthesize images of An. gambiae s. 1. aquatic habitats. The georeferenced data was then transferred into a SAS/GIS module which was used to generate univariate statistics, correlations, distributions, and to create global autocorrelation statistics. A local autocorrelation index was then constructed using spatial covariance parameters （ i. e. , Moran＇s Indices）. The model generated a robust subset of selected eigenvectors which identified spatial interaction between the sampled variables. Coefficient estimates were then used to define expectations for prior distributions in a Bayesian estimation matrix, which revealed that the covariate depth of habitat was a significant predictor variable, positively associated with An. gambiae s. 1. aquatic habitats. A specific subset of eigenvectors from a transformed spatial link matrix can capture dependencies among the disturbances of a spatial regression model generated using An. gambiae s. 1. aquatic habitat predictor variables.