In social epidemiology, a geographic neighborhood or cluster is viewed as an important determinant of health behaviors, mediators, and outcomes. One may be interested in the effects of measured or unmeasured neighborhood characteristics or in turn on individual effects that have been disentangled from neighborhood effects. Analyses of nationally representative surveys, such as the National Health Interview Survey, provide a means of estimating these effects. This project will develop statistical methods that can account for the complex sampling design of such surveys at the same time as disentangling individual effects from neighborhood effects. These methods will be applied to analyze data from the National Health Interview Survey. Furthermore, in global health and several other fields, community randomized trials with complex sampling designs are used to estimate the effect of one or more interventions versus a standard or control condition. Scientific interest often focuses on the relative effects of community-level adherence to the intervention on individual-level outcomes. The project also will develop statistical methods that can account for the complex sampling design of such trials while simultaneously extracting the effect of intervention adherence, as opposed to that of intervention intention on individual-level outcomes. These methods will be applied to analyze data from randomized trials designed to study effects of school-level sanitation, water safety, and hygiene on individual education outcomes.
This project involves collaborative research across the fields of biostatistics, social epidemiology, and global health. The research will advance statistical methodology as well as improve the capability of researchers in social epidemiology, global health, and other fields to address important scientific questions. Disentangling individual-level effects from neighborhood-level effects will be useful in understanding the relative roles of the individual versus society and environment in health behaviors and outcomes, which will be useful in designing interventions. Going beyond simple comparisons of treated and untreated individuals in randomized clinical trials and estimating the effects of community-level intervention adherence on individual-level outcomes will further understanding of the effects of interventions in global health and other fields. Illustrating and communicating the new statistical methods for joint estimation of individual-level and neighborhood-level effects on outcomes using nationally representative surveys will provide the federal agencies that sponsor these surveys with enhanced options for creating public-use datasets to facilitate these analyses. The project is supported by the Methodology, Measurement, and Statistics Program and a consortium of federal statistical agencies as part of a joint activity to support research on survey and statistical methodology.