Spatial aggregation as a means to improve attribute reliability

Attributes of areal units are often estimates derived from survey samples. Estimates of these attributes with large standard errors (SEs) discount the confidence and validity of spatial analytical results. Large SE for estimates of enumeration units are often the results of small sample sizes in areal units and imply unreliable attribute values. One way to suppress error in attributes is to merge areal units to raise sample size. Traditional regionalization methods serve this purpose, but may unnecessarily alter the geography of the study area. We propose an interactive-heuristic aggregation approach to assist analysts in selecting and merging only units with SEs larger than acceptable levels while preserving the original geography and data as much as possible. Results of this approach and a recent automated optimization method are comparable. Both methods successfully lower the SEs in attribute data, but the interactive approach flexibly adjusts the importance levels of different aggregation criteria across areal units, thus offering a high degree of transparency in the aggregation process. The interactive approach also incorporates subjective and local knowledge of neighborhoods in selecting areal units for aggregation.