HACKING SPATIAL DATA: AN EXAMPLE OF AGGREGATION PROBLEMS
Many applications using spatially aggregated data tend to treat the spatial units as given. For example, in the United States, analyses using the social and economic data often rely on the existing and fixed spatial units of census blocks or tracts. However, these spatial units are often aggregated arbitrarily. It is therefore important to ask this question: what if the spatial units are aggregated differently? Will the results obtained using the existing units still hold? This paper addresses questions like these. We first develop a search algorithm that can be used to find alternative aggregations with relatively equal total populations among the aggregated units. Then a number of experiments are conducted to test the algorithm and to demonstrate how alternative aggregations will affect the analysis. These experiments clearly suggest the significant effects of spatial aggregation on the analysis results.