How Far Is Too Far? - A Statistical Approach to Context-contingent Proximity Modeling

Proximity is a fundamental concept in any comprehensive ontology of space (Worboys 2001). The provision of a context-contingent translation mechanism between linguistic proximity measures (e.g. “near”, “far”) and metric distance measures is an important topic in current GIS research. After a discussion of context factors that mediate the relationship between linguistic and metric distance measures, we present a statistical approach, Ordered Logit Regression, to the context-contingent proximity modeling. The approach can predict proximity given the corresponding metric distance and context variables. An empirical case study with human subjects is carried out using this statistical approach. Interpretation and predictive accuracy of the empirical case study are discussed.