Obesity and other chronic diseases are becoming more prevalent in affluent countries such as Australia. Researchers are trying to understand and combat this trend. One related growing stream of research explores the role of the built environment and transport system on an individual's weight. However, results from many studies conducted have been contradictory. A primary cause of these contradictions is due to how neighborhood areas are defined, which directly affects how the built environment variables are calculated in geographic information systems. The potential impacts on regression analysis resulting from different data aggregation methods are well documented in spatial studies, geography, and regional planning fields, and the problem is primarily referred to as the modifiable aerial unit problem. In this paper, the focus is on reducing the error caused by the modifiable aerial unit problem by introducing a new data aggregation method. Individual health and lifestyle data are obtained from the survey of households, income, and labor dynamics in Australia, and the relationship between the built environment and obesity is evaluated by using a discrete choice model. The proposed aggregation method is evaluated across three spatial scales and compared against a conventional data aggregation method (i.e., using predefined administrative boundaries such as census tracts). The results reveal a stronger relationship between land use variables and obesity when the proposed aggregation method is implemented. This paper is relevant primarily to researchers because it provides an improved aggregation method to deal with some privacy restrictions of surveys. It is also relevant to practitioners and policy makers by its quantification of the association between specific built environment variables and obesity.