The Modifiable Areal Unit Problem in Multivariate Statistical Analysis

Abstract

In this paper the examination of the modifiable areal unit problem is extended into multivariate statistical analysis. In an investigation of the parameter estimates from a multiple linear regression model and a multiple logit regression model, conclusions are drawn about the sensitivity of such estimates to variations in scale and zoning systems. The modifiable areal unit problem is shown to be essentially unpredictable in its intensity and effects in multivariate statistical analysis and is therefore a much greater problem than in univariate or bivariate analysis. The results of this analysis are rather depressing in that they provide strong evidence of the unreliability of any multivariate analysis undertaken with data from areal units. Given that such analyses can only be expected to increase with the imminent availability of new census data both in the United Kingdom and in the USA, and the current proliferation of GIS (geographical information system) technology which permits even more access to aggregated data, this paper serves as a topical warning.

Spatial scale effects in environmental risk-factor modelling for diseases

“…Justifications for such choices are seldom provided in the literature. It is indeed difficult to determine a spatial extent that is agreeable for everyone but on the other hand, many landscape ecological studies have repeatedly shown that spatial extents do determine the range of patterns and processes that can be detected on a landscape, and have cautioned researchers of the uncertainties associated with spatial extent changes (Turner et al, 1989;Fotheringham et al, 1991;Wu, 2004). Space and therefore land cover/land use are continuous phenomena and drawing any discrete boundaries over them to extract meaningful data will introduce complications.…”

“…total area, percent cover) will differ as well. It has been long recognised that among some of the methodological problems encountered when using geospatial analysis in general, including its application in spatial epidemiology, the most common are those that are associated with spatial scales (Fotheringham and Wong, 1991;Sexton, 2008). The term "scale" has been used in multiple contexts to mean different concepts and their meanings are not interchangeable.…”

Making Space for Theory: The Challenges of Theorizing Space and Place for Spatial Analysis in Criminology

“…These issues have important implications for ecological studies of crime as multivariate statistical analyses can be sensitive to variations in scale and zoning systems, leading to highly unreliable results (e.g., Fotheringham and Wong 1991). The problems posed for statistical inference from MAUP have led some to conclude that all methods whose results depend on areal units should be discarded and techniques independent of areal units should be used (e.g., Tobler 1989; see also Openshaw and Taylor 1981;Openshaw 1984;Fotheringham 1989;Fotheringham and Wong 1991;Fotheringham and Rogerson 1993).…”

“…The problems posed for statistical inference from MAUP have led some to conclude that all methods whose results depend on areal units should be discarded and techniques independent of areal units should be used (e.g., Tobler 1989; see also Openshaw and Taylor 1981;Openshaw 1984;Fotheringham 1989;Fotheringham and Wong 1991;Fotheringham and Rogerson 1993). Gridbased models have also been advocated in spatial analysis to avoid the use of inconsistently sized areal units but the issues of the choice of grid size and the associated level of aggregated information remains.…”

Deprivation, Healthcare Accessibility and Satisfaction: Geographical Context and Scale Implications

“…The phenomenon of the modifiable areal unit problem (MAUP) usually causes the results of statistical analyses to differ according to the scale and size of the spatial reporting units (Briant et al 2010;Openshaw 1984). The effects of MAUP have been analyzed using several approaches, generally based on correlation and regression methods (Flowerdew 2011;Fotheringham and Wong 1991;Openshaw 1984;Pietrzak 2014;Soobader et al 2001), and these kinds of analyses have typically been applied to census indicators (Flowerdew 2011;Soobader et al 2001).…”