next up previous
Next: Spatial econometrics and lattice Up: Exploratory spatial data analysis Previous: Local indicators of spatial

The Modifiable Areal Unit Problem (MAUP)

Having outlined the MAUP above, it remains here to indicate progress in addressing and in part resolving the issues involved. Arbia (1989) made a major contribution by studying in depth a range of links between the presence of spatial autocorrelation and the MAUP; until that time most analysts had chosen to sidestep Openshaw and Taylor's (1979) potentially devastating finding that the results of statistical analysis of data for spatial zones could be varied at will by changing the zonal boundaries. The problem includes two parts, the problem of scale, involving the aggregation of smaller units into larger ones, and the problem of alternative allocations of component spatial units to zones, also known as gerrymandering.

A further positive contribution was made by Fotheringham and Wong (1991), followed up by Amrhein (1995), Fisher and Langford (1995), Amrhein and Reynolds (1996), and Morphet (1997). Openshaw (1996) summaries many of the technologies now available for choosing zoning systems to optimize results. Perhaps the most active group of recent publications has resulted from collaboration between social statisticians experienced in complex survey design and geographers, including Holt, Steel, Tranmer, and Wrigley (1996) and Holt, Steel, and Tranmer (1996), and Wrigley et al. (1996). Focusing closely on the scale and zoning effects, they conclude that the use of well chosen grouping variables to adjust the area-level results may yield reliable estimates of underlying individual-level relationships, thus providing at least a partial solution to the MAUP with respect to the ``ecological fallacy'', the drawing of individual-level inferences based on area-level analyses.



Roger Bivand
Fri Mar 5 08:30:34 CET 1999