While distinct methods, spatial statistics and GIS rely on each other in important ways. GIS software and tools can be used to create covariates for inclusion in statistical models and to visualize the output from statistical models. Spatial statistics provides rigorous modelling and inference techniques for drawing conclusions from geographically-indexed data. In particular, spatial statistics provides a body of methods for spatial smoothing and for accounting for nonspatial covariates in estimating spatial surfaces. While implementations of some statistical models, such as kriging, are included in standard GIS software, many researchers prefer using statistical software such as R and S-Plus. In turn, techniques for working with geographic data, including mapping and visualization, are less developed in statistical software than in GIS software. Former HSPH assistant professor Chris Paciorek has written an overview of fitting spatial surfaces, in which he describes some of the advantages of statistical methods implemented in statistical software over the implementation of kriging in GIS software.
The Department of Biostatistics at Harvard School of Public Health has an active research group in the area of environmental statistics, including methodological and applied research in spatial statistics. Faculty members Brent Coull, Yi Li, Xihong Lin, Christopher Barr and Francesca Dominici are all involved in research projects involving spatial statistics. Dr. Coull’s work focuses on both point-level and areal data, with application to air pollution exposure and epidemiology and to disease mapping. Dr. Li’s work focuses on spatial correlation in survival analysis. In addition, Marcello Pagano, also a faculty member in the department, leads an active group working on disease surveillance, which includes analysis of spatial and spatio-temporal data. His collaborators include adjunct faculty Al Ozonoff and Laura Forsberg White.
Last modified 1/10/2013.