Statistical Methods for Environmental Social Science

最近，公众和研究人员对人类与环境之间的因果关系产生了极大的兴趣。这种兴趣主要针对回答环境正义的问题，即确定某些社会人口亚群是否承担了某些环境危害的不适当份额。促进这种兴趣的是同时出现的地理信息系统(GISS)，这是一种复杂的计算机程序，可以在一个共同的研究区域内对多个数据源进行分层。然而，这类程序目前没有统计推断的设施，这对于确定数据是否提供了不公正的“重要”证据，或者调查结果是否完全是偶然发生的是必要的。本项目的重点是开发必要的统计分层建模方法，以改进对这些数据集的分析，以及相应的用户友好的计算工具，以使非专家统计支持人员能够使用这些方法。该项目将集中在至少三个环境社会科学应用领域。首先，将考虑处理未对齐的面状数据的统计方法(这类数据很容易由GISS绘制)。在这里，这些方法必须适应在错位的区域边界(例如，邮政编码和人口普查区域)以及随时间演变的边界上收集的数据。其次，这些方法将扩展到数据不仅未对齐，而且是不同类型的情况。例如，一个变量可能只能作为区域聚合(例如，邮政编码外国出生的百分比)，而另一个变量只能在空间的某些点(例如，一组固定监测站的每日颗粒物水平)提供。最后，这些方法将进一步扩展，纳入解决多个相互冲突的优先事项和目标的正式工具。实现环境公平往往涉及在效率(即最大限度地扩大成本和收益之间的总体差异)和公平(即公平地分配这些成本和收益)之间进行权衡。为社会带来的好处将包括改进的环境正义评估、在不同尺度上处理数据的系统方法，以及解决相互冲突的分析优先事项和目标的实用战略。这些应该是公共政策制定者感兴趣和有用的，因为他们正在努力解决清理现有环境危害以及公平地选址未来可能存在危险的设施的问题。

Recently, there has been an explosion of both public and research interest in the cause-and-effect relationships between humans and their environment. Much of this interest has been directed at answering questions of environmental justice, i.e., determining whether certain sociodemographic subgroups bear an undue share of certain environmental hazards. Helping facilitate this interest has been the concurrent emergence of geographic information systems (GISs), sophisticated computer programs for "layering" multiple data sources over a common study area. However, such programs at present have no facility for statistical inference, which is necessary for determining whether the data offer "significant" evidence of injustice, or whether the findings could just as well have arisen completely by chance.This project focuses on developing necessary statistical hierarchical modeling methods for the improved analysis of such data sets, as well as corresponding user-friendly computing tools to enable use of the methods by non-expert statistical support staff. The project will focus on at least three environmental social science application areas. First, statistical methods for handling misaligned areal data (of the sort so easily mapped by GISs) will be considered. Here the methods must accommodate data aggregated over misaligned regional boundaries (say, zip codes and census tracts), as well as boundaries which evolve over time. Second, the methods will be extended to the case where data are not only misaligned, but of different types. For example, one variable might be available only as a regional aggregate (say, percent foreign-born by zip code), while another is available only at certain points in space (say, daily particulate matter levels at a collection of fixed monitoring stations). Finally, the methods will be further extended to incorporate formal tools for resolving multiple and conflicting priorities and goals. Achieving environmental equity often involves making tradeoffs between efficiency (i.e., maximizing the overall difference between costs and benefits) and equity (i.e., an evenhanded distribution of these costs and benefits). Benefits accruing to society will include improved environmental justice assessments, a systematic approach for handling data at different scales, and practical strategies for resolving conflicting analytic priorities and goals. These should be of interest and use to public policymakers as they grapple with the problems of cleaning up existing environmental hazards, as well as equitably siting future potentially hazardous facilities.