A New Statistical Method for Analyzing Point Collocations

This article proposes a new method for analyzing the collocation of spatial objects represented as points. We consider the methodological aspect of point collocation. Our focus is on the collocation of more than two types of points, such as the collocation of clothing, accessory, and shoe stores, as well as that of hamburger, sandwich, and pizza stores. Although such collocations are often discussed in geography, analytical methods have not yet been fully developed. To fill the research gap, we develop a new statistical method for detecting the collocations of more than two types of points. We tested the validity of the method through computational experiments and applications to the collocation analysis of restaurants in the Tokyo metropolitan area. The former supported the statistical power of the proposed method, whereas the latter showed its efficiency in detecting strong collocations and provided interesting empirical findings.