Capturing spatial co-location patterns—subsets of two or more types of events that are geographically close—is one of the primary interests in spatial analysis. For example, different types of retail outlets and service firms are clustered together in central districts of cities across the world to take advantage of spatial agglomeration. To analyze such spatial phenomena, the cross K function has been widely used. However, this method and other existing methods are likely to be unsuitable for analyzing co-locations in a micro-scale space due to some limitations.
To precisely analyze the micro-scale co-location, this project develops a new statistical measure named network dual K function. Compared to the ordinary cross K function, the proposed method has distinctive features such as assuming network-constrained space and using an exact statistical formula for a test.
This dissertation proposes a set of three methods of network dual K function: global, local, and incremental methods. The proposed methods are applied to various types of stores in trendy districts in Tokyo to demonstrate the usefulness of the methods for studies on economic geography, and to improve our understanding of urban agglomeration. In addition, this research project highlights the significance of conducting network spatial analysis.