Selecting Spatial Scale of Area-Level Covariates in Regression Models

SELECTING SPATIAL SCALE OF AREA-LEVEL COVARIATES IN REGRESSION MODELS By Lauren P. Grant, Ph.D. A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy at Virginia Commonwealth University. Virginia Commonwealth University, 2016 Director: David C. Wheeler, Ph.D., MPH Assistant Professor, Department of Biostatistics Director: Chris Gennings, Ph.D. Research Professor, Department of Preventive Medicine Icahn School of Medicine at Mount Sinai In epidemiological and environmental studies, investigators are often interested in the contextual or area-level effects that are associated with a specific health outcome. Area-based covariates are typically available at multiple spatial scales (i.e., areal units or buffer distances). Studies have found that the level of association between an area-level covariate and an outcome can vary depending on the spatial scale (SS) of a particular covariate. However, covariates used in regression models are customarily modeled at the same spatial unit. In this dissertation, we developed four SS model selection algorithms that select the best spatial scale for each area-level covariate. The SS forward stepwise, SS incremental forward stagewise, SS least angle regression (LARS), and SS lasso algorithms allow for the selection of different area-level