Conflating Survey Data into Sociocultural Indicator Maps
Abstract
This report presents a methodology of mapping population-centric social, infrastructural, and environmental metrics at neighborhood scale. This methodology extends traditional survey analysis methods to create cartographic products useful in agent-based modeling and geographic information analysis. It utilizes and synthesizes survey microdata, sub-upazila attributes, land-use information, and ground-truth locations of attributes to create neighborhood-scale multi-attribute maps. Monte Carlo methods are used to combine any number of survey responses to stochastically weight survey cases and to simulate survey-case locations in a study area. Through these methods, known errors from each input source can be retained. By keeping the individual survey case as the atomic unit of data representation, this methodology ensures that important covariates are retained and that ecological inference fallacy is eliminated. These techniques are demonstrated using data and output maps for Chittagong Division, Bangladesh. The results provide a population-centric understanding of many social, infrastructural, and environmental metrics desired in humanitarian aid and disaster relief planning and operations wherever long-term familiarity is lacking. Of critical importance is that the resulting products have easy-to-use explicit representation of the errors and uncertainties for each input source via the automatically generated summary statistics created at the applications geographic scale.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2018
- Accession Number
- AD1081477
Entities
People
- Carey L. Baxter
- Charles R. Ehlschlaeger
- Dandong Yin
- David A. Brown
- Ellen R. Hartman
- James D. Westervelt
- Jeffrey A. Burkhalter
- Marina V Drigo
- Matthew D. Hiett
- Natalie R. Myers
- Robert C. Lozar
- Scott A. Tweddale
- Yizhao Gao