User:Geo arbo/GES 679/assignment 01

=Recognizing MAUP in Research=

The Modifiable Areal Unit Problem (MAUP) is an issue in many analyses that attempt to find patterns of clustering in aggregated data. The problem arises because aggregated data are spatially delimited by boundaries. Analytical results can be distorted if the dimensions of boundary units and the spatial relationships of sets of boundary units are not chosen carefully. Successful cluster analyses are based on areal units and areal unit arrangements that are appropriate for the data being studied. It is good practice to use spatial statistical tools to evaluate data with the goal of minimizing areal unit problems.



Cluster analysis methods were used by Bishop and Cushing 2008 in an analysis of United States population migration trends. The study period was 1948-2004. The areal units used were counties. Counties were divided into classes: counties with close Presidential races and counties where Presidential candidates won by landslides (winning margins of at least 20%). The study tracked migration of Democratic and Republican voters into landslide counties. Migration patterns between low population (rural) and high population (urban) counties were tracked. Patterns in relationships between party membership, education, income and religious practices were described in the article.

A goal of the analysis was to discover whether people with similar politics tended to move to counties where landslide victories for their political party had occurred. Figure 1 illustrates one of Bishop and Cushing’s findings: the likelihood that average Americans would live in ‘Landslide’ counties with neighbors who had similar politics increased from roughly 26% in the year 1976 to 50% the year 2004.

A monograph written by Openshaw in 1984 reviewed and discussed the Modifiable Areal Unit Problem (MAUP) and associated spatial analytical issues. He provided examples of how different choices of analytical areal units skewed results. Openshaw argued that improved methods of minimizing analytical artifacts of poorly chosen spatial analytical units were possible and desirable. If counties were not the most appropriate units for Bishop and Cushing’s study, there could be problems with their conclusions.

The article related to the GES 679 13-Sep-2011 lecture in the following ways. Census data were aggregated by counties and the lecture noted that aggregation can "amplify or suppress existing patterns" as well as create "new patterns that are artifacts of the aggregation process". Lecture statements regarding bias introduced by variation in zonal units and scale may be relevant to Bishop and Cushing’s analyses derived by aggregating Census data into county units and geographic arrangements.

For example,there are 62 counties in New York state and year 2009 county population estimates ranged from a minimum of 4,923 to ranged from a maximum of 2,567,098. The maximum New York county population is well over 500 time larger than the minimum county population. Nationally, there is also a wide range in county population numbers. [Chttp://www.census.gov/geo/www/cen_tract.html Census tracts] usually have between 2,500 and 8,000 persons. If Bishop and Cushing had chosen to use census tracts, which have comparable population values, as areal units instead of counties, it is possible that there results would have been different.