Spatial Statistics for Alcohol Epidemiology

描述(由申请人提供)：我们提出了一个跨学科的研究计划，解决空间分析方法在分析现有的酒精流行病学数据集，来自德克萨斯州的休斯顿和奥斯汀，以及加利福尼亚州的萨克拉门托。该项目通过一个由预防科学家和空间统计学家组成的团队，为该领域的长期研究合作奠定了基础。具体目标包括三个方法论目标：1.定义和应用具有空间相关性的层次化广义线性模型。我们将开发和应用具有相关误差的线性、二项和泊松回归模型来模拟数据集中计数和比率的空间模式。具体工作包括在各种空间尺度上估计当地酒精销售点密度、非法药物使用和暴力后果之间的联系。2.定义和应用酒精流行病学的变系数模型。我们建议开发和比较统计方法与位置与风险因素的相互作用，其中相关性在研究区域内不同。应用包括估计研究区域内酒精之间关联性的局部变异。应用包括在示例数据集中估计酒精出口密度、非法药物使用和暴力收入之间的关联的局部差异。3.定义和应用网络位置分布的空间分析方法。我们将开发和应用“网络空间”的分析方法，其中距离是通过网络旅行而不是直线来定义的。这项工作以新的和非标准的方式将地理信息系统与空间统计联系起来，并能够在奥斯汀数据中对酒精销售点的空间模式以及这些模式与犯罪活动模式之间的联系进行新的量化。 总结。我们的具体目标涉及方法学发展的三个领域，但每个领域都提供了解决酒精流行病学中特定假设的工具。这些方法可以调查从“出口密度是否影响暴力发生率？”等具体问题。到“不同人口群体在暴露和/或健康风险方面是否存在差异？”在一定的地理范围内。

DESCRIPTION (provided by applicant): We propose an interdisciplinary research program addressing spatial analytic methods in the analysis of existing alcohol epidemiology data sets from Houston and Austin, Texas, and Sacramento, California. The project builds the foundation for long-term research collaborations in this area via a team of prevention scientists and spatial statisticians. The Specific Aims involve three methodological goals: 1. Define and apply hierarchical generalized linear models with spatial correlation. We will develop and apply linear, binomial, and Poisson regression models with correlated errors to model spatial patterns in counts and rates in the data sets. Particular efforts include estimation of associations between local alcohol outlet density, illegal drug use and violent outcomes across a variety of spatial scales. 2. Define and apply variable coefficient models for alcohol epidemiology. We propose development and comparison of statistical methods with location-by-risk factor interactions where associations vary across the study area. Applications include estimation of local variations in association between alcohol across the study area. Applications include estimation of local variations in association between alcohol outlet density, illegal drug use and violent incomes in the example data sets. 3. Define and apply methods for the spatial analysis of location distributions on a network. We will develop and apply analytic approaches on "network space" where distances are defined via network travel rather than straight lines. This work links geographic information systems (GISs) with spatial statistics in new and non-standard ways and enables novel quantification of spatial patterns of alcohol outlets and the associations between these patterns and patterns of criminal activity in the Austin data. Summary. Our specific aims address three areas of methodological development but each provides tools addressing specific hypotheses in alcohol epidemiology. The methods enable investigation of specific questions ranging from "Do outlet densities affect violence rates?" to "Are there disparities in exposures and/or health risks for different demographic groups?" within a geographic context.