Detecting and analyzing patterns in spatial data

空间分布的感知模式是环境病因学中最具挑衅性的线索之一。根据人口动态数据制作的地图很容易获得，这提高了公众对疾病模式差异的兴趣和认识。另一方面，流行病学家对这类数据的方法局限性感到不安，许多人认为随访是徒劳的。以前对集群和生态（群体水平）分析的方法论批评主要是理论上的，并且不知道在实践中实际发生的问题的程度或原因。回答这些问题需要详细的个人层面的空间流行病学数据（“金标准”）进行比较，但这些数据很少可用。来自Cape Code的关于癌症（来自当前SBRP中心）和生殖/发育影响（项目1，本申请）的地理编码数据提供了这样的资源。两种新方法，一种用于识别疾病“热点”（由我们小组开发），另一种用于生态分析（King的“EI方法”），显示出用于环境流行病学的前景。在本项目中，我们将继续开发自适应速率平滑方法，以显示当地的疾病过度，增加协变量调整的能力;并开发方法学工具来分析和可视化潜在的生态数据问题。我们将这些方法应用于开普敦的数据，研究空间混杂所造成的聚类程度，并比较生态推理与个人层面的开普敦数据使用传统的和新的生态方法的推断。

Perceived patterns in spatial distributions are among the most provocative clues to environmental etiologies. The easy availability of maps made from vital events data have heightened public interest and awareness of these differences in disease patterns. Epidemiologists, on the other hand, are uneasy about the methodological limitations of these kinds of data and many believe follow-up is unproductive. Previous methodological critiques of cluster and ecological (group-level) analyzes are largely theoretical, and it not known to what degree problems actually occur in practice or why. Answering these questions requires detailed individual- level spatial epidemiological data ("the gold standard") for comparison, but these data are rarely available. Geographically coded data from Cape Code on cancer (from the current SBRP center) and reproductive/developmental effects (Project 1, this application) provide such a resource. Two new methods, one for identifying disease "hotspots" (developed by our group), and one for ecological analysis (King's "EI method") show promise for use in environmental epidemiology. In this project we will continue development of the adaptive rate smooth method for showing local disease excesses, adding the ability to adjust for covariates; and develop methodological tools to analyze and visualize the potential for problems form ecologic data. We will apply these methods to the Cape data, examining the degree of clustering caused by spatial confounding and compare ecological inference with individual-level inference of the Cape data using both conventional and novel ecological methods.