Centre for Statistical & Survey Methodology Working Paper Series

Publication Date



This paper aims to explore the modifiable areal unit problem (MAUP) from a spatial analysis perspective. Using different scales and zoning on a particular set of spatial data may lead to problems in interpreting the results. Pawitan et Steel (2006) explored the bias associated with aggregate level semivariogram analysis in comparison with the corresponding individual level analysis. Minot et Baulch (2005a) discussed some consequences of using aggregated level data in poverty mapping, which may affect the validity of the output. Tagashira et Okabe (2002) investigated the consequences of the MAUP in developing regression models for spatially aggregated data.

TheMAUP will be examined theoretically and empirically based on census data, as well as simulations. Discussion focuses on the expectation of differences between analysis undertaken at two different aggregation levels. We introduce a semivariogram model for aggregated data, and explore relationships using the model at different levels of aggregation.