We provide advice on data analysis for neighborhood health effects studies and are happy to partner with an outside group to collaborate on data analysis. The major analytical issues that need to be addressed in these studies are: 1) the non-independence of observations for study subjects living in the same neighborhood or spatial unit; and 2) the potential for spatial auto-correlation and potential spill-over effects between neighboring spatial units. We have extensive experience using multi-level modeling, generalized estimating equation and cluster-robust based methods for analyzing neighborhood health effects data.
Another common data analysis issue is how to analyze the multitude of measures that might be available for a particular construct, this is a common concern for constructs like neighborhood socio-economic status where there are many possible indicator variables (e.g. poverty rate, median household income, educational attainment…). We have experience working with standard indexes such as, the Neighborhood Deprivation Index, neighborhood walkability measures and neighborhood disorder measures. We also have experience developing our own indexes measuring neighborhood contextual constructs using factor analysis and clustering methods.
A similar issue arises when measuring the retail environment. We have extensive experience using business listing data (Dun & Bradstreet, NETS, InfoUSA) to measure neighborhood food and physical activity environments and have developed several summary measures describing neighborhood food environments.
We have also developed a simulation platform to test potential data analysis strategies. The platform is based on the New York City Community Health Survey data and uses Zip codes as the neighborhood definition.
Built Environment and Health Research Group 