The Uncertain Geographic Context Problem

Abstract

Any study that examines the effects of area-based attributes on individual behaviors or outcomes faces another fundamental methodological problem besides the modifiable areal unit problem (MAUP). It is the problem that results about these effects can be affected by how contextual units or neighborhoods are geographically delineated and the extent to which these areal units deviate from the true geographic context. The problem arises because of the spatial uncertainty in the actual areas that exert the contextual influences under study and the temporal uncertainty in the timing and duration in which individuals experienced these contextual influences. Using neighborhood effects and environmental health research as a point of departure, this article clarifies the nature and sources of this problem, which is referred to as the uncertain geographic context problem (UGCoP). It highlights some of the inferential errors that the UGCoP might cause and discusses some means for mitigating the problem. It reviews recent studies to show that both contextual variables and research findings are sensitive to different delineations of contextual units. The article argues that the UGCoP is a problem as fundamental as the MAUP but is a different kind of problem. Future research needs to pay explicit attention to its potential confounding effects on research results and to methods for mitigating the problem.

En cualquier estudio que examine los efectos que tienen atributos basados en área sobre conductas o logros individuales se tiene que enfrentar un problema metodológico fundamental aparte del problema de la unidad espacial modificable (PUEM). El inconveniente es que los resultados acerca de estos efectos pueden afectarse por la manera como las unidades contextuales o vecindarios sean delineados geográficamente y el grado en que dichas unidades espaciales se apartan del contexto geográfico verdadero. El problema surge debido a la incertidumbre espacial en las áreas reales que ejercen las influencias contextuales bajo estudio y la incertidumbre temporal presente en el cronograma y duración con que los individuos experimentaron estas influencias contextuales. Utilizando los efectos de vecindad y la investigación sobre salubridad ambiental como punto de partida, este artículo hace claridad sobre la naturaleza y orígenes de este problema, que se denomina problema de incertidumbre del contexto geográfico (PICoG). Se destacan algunos de los errores de inferencia que el PICoG podría ocasionar y se discuten algunos medios para mitigar el problema. En el artículo se revisan también estudios recientes para mostrar que tanto las variables contextuales como los hallazgos de la investigación son sensibles a las diferentes demarcaciones de las unidades contextuales. En el artículo se arguye que el PICoG es un problema tan fundamental como el PUEM, aunque se trata de un tipo diferente de problema. La futura investigación sobre el particular deberá poner atención explícita a los potenciales efectos desorientadores de los resultados que se logren y a los métodos para mitigar el problema.

Key Words:

• contextual uncertainty
• environmental health
• neighborhood effects
• uncertain geographic context problem
• UGCoP

关键词:

• 情景的不确定性
• 环境卫生
• 邻里效应
• 地理环境不确定性问题
• UGCoP。

Palabras clave:

• incertidumbre contextual
• salubridad ambiental
• efectos de vecindad
• problema de incertidumbre del contexto geográfico

Acknowledgments

I thank Antoinette WinklerPrins for handling the blind review process and making editorial decisions for this article (including the abstract). Her helpful suggestions and the thoughtful comments of three anonymous reviewers have helped improve the article considerably. This article was written while I was supported by the following grants: NSF BCS-0729466, NCI R21 CA129907, and NIDA R01DA025415-01.

Notes

1. Two important qualifications of the article's focus on health studies are in order. First, arguments in this article mainly apply to health studies that are based on ecological designs because other research designs are not primarily concerned with identifying contextual influences from geographic factors and processes (and thus do not need to explicitly delineate contextual areas). Second, discussion in this article is relevant mainly to studies in which area-based contextual variables (e.g., neighborhood poverty) are used to explain or predict individual health behaviors or outcomes. An important goal of many health studies, however, is to identify at-risk populations or areas where the health outcomes are significantly worse than expected. Given their analytical focus on the relationship between area-based contextual variables and area-based outcome variables (e.g., low birthweight rates of census tracts), using conventional administrative units like census tracts in this kind of study is needed and is often the only viable option.

2. A major concern with collecting GPS data in health research is participants’ privacy and data confidentiality, because it could be possible to identify a person's identity through reverse geocoding if data are not handled carefully. In countries with strict human subject protection regulations (such as the United States), all persons involved in collecting and analyzing the data are required to go through rigorous human subject protection training and be certified before any involvement in research activities. They are obliged legally and ethically to protect participants’ privacy and data confidentiality, and all research procedures (including recruitment, informed consent, data analysis, and dissemination of results) require prior approval from and are closely monitored through continuing review by their institutional or ethical review boards. For instance, in the Appalachian smokeless tobacco usage study, all data were deidentified before being incorporated into the database and no one handling those data or seeing them by chance will be able to identify any participant. Further, no maps or displays of the home or activity sites visited by participants or their daily paths can be printed or disseminated. The GPS data cannot be used in any form other than for deriving activity spaces and related measures or generating aggregate statistical results. In countries without strict regulations and in situations where people provide information without knowing or agreeing to its use for research purposes (e.g., social networking Web sites such as Facebook Places, Twitter, and Foursquare), there might be no informed consent and human subject protection protocol and issues of privacy violation can be a serious concern. It is not clear how the use of location data can be justified with respect to the norms of human subject protection in these contexts.