|
Abstract:
|
Statistical independence is a fundamental property underlying much of inferential theory, yet seems rather the exception in the practical world around us. Just as many time series exhibit temporal dependencies, so also many cross-sectional data sets exhibit spatial dependencies. There is a developed literature on the identification and weighting of spatial neighbors, and these measures form a basis for capturing the effects of spatial dependencies and clusterings. Unfortunately, the choice among weighting schemes often is rather arbitrary and ad hoc, even though the consequences of different choices can be substantial. This paper surveys a wide range of spatial weight functions, demonstrating the variation in results from different choices. The focus is on practical illustrations of effects, using social and economic data collected across contiguous zip codes in a metropolitan area.
|
