Prof. Bob Haining has been the professor of Human Geography at Department of Geography, University of Cambridge since 2000. He is the co-founder of the Spatial Econometrics Association and is an associate editor of several prominent Geographical and interdisciplinary journals including the Journal of Geographical Systems, Geographical Analysis, Spatial and Spatio-Temporal Epidemiology and Computational Statistics. For more than thirty years he has made significant contributions to spatial analysis with applications in the areas of spatial epidemiology and health services research, the geography of crime and economic geography. He has published widely, including two popular books: Spatial Data Analysis in the Social and Environmental Sciences (1990) and Spatial Data Analysis: Theory and methods (2003), and articles in over 20 different academic journals in areas that include geography, statistics, economics and epidemiology.
Spatial precision and statistical precision: can we have the best of both worlds?
Modern digital technology enables us to collect, store and map social and environmental data in fine geographical detail. Such spatial precision is of considerable benefit in helping us to observe and track change at the small area level in for example disease risk, crime risk and environmental risks. This in turn has the potential to enable us to better understand the processes at work and to respond to threats to life and well-being.
However to be able to make the best use of such spatially fine-grained data we often need to be able to obtain reliable small area level estimates of attribute properties. But in many instances estimates for small geographical areas suffer from the “small number problem” and a lack of statistical precision. Estimator error in such circumstances may be unacceptably large thereby undermining the benefits that can be derived from such data.
I shall look at ways geographical information scientists and spatial statisticians have tried to tackle this problem of spatial data quality in order to have the “best of both worlds” – spatial precision and statistical precision.