Minimizing effects of scale distortion for spatially grouped census data using rough sets

摘要

Census data has been widely used for community evaluation based on demographic and socioeconomic variables. However, the analysis is typically associated with specific areal units and the results often change when the size of the census configuration changes leading to scale distortions. Various approaches such as optimal zoning systems and multivariate statistical analysis have been developed to address the scale problem. But limitations in these approaches have led to the use of non-statistical methods to tackle the scale problem. This study combines a non-statistical method with descriptive statistical measures to develop a rough sets approach to constructing a census-based deprivation index (DI) and to determine its relationship to a recent immigrant population using the 2001 Canadian census. Application of the approach in the Greater Vancouver Regional District shows that rough sets can stabilize relationships for spatially grouped census data by minimizing scale distortions. Scale sensitivity measures are also estimated to translate DI relationships across three census configurations. The rough sets approach is suitable for areal data analysis because it is resistant to nonlinearity, outliers, and assumes no prior relationship between variables.