Premium
2.15 or Not 2.15? An Historical‐Analytical Inquiry into the Nearest‐Neighbor Statistic
Author(s) -
Philo Chris,
Philo Peter
Publication year - 2022
Publication title -
geographical analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.773
H-Index - 65
eISSN - 1538-4632
pISSN - 0016-7363
DOI - 10.1111/gean.12284
Subject(s) - statistic , randomness , sophistication , point (geometry) , completeness (order theory) , chart , k nearest neighbors algorithm , statistics , mathematics , computer science , sociology , artificial intelligence , social science , mathematical analysis , geometry
Nearest‐neighbor analysis (NNA)—a method for assessing the degree to which a spatial point pattern departs from randomness in the direction of being either clustered or regular—was imported into academic geography from an article published in 1954 by ecologists Clark and Evans. In its simplest form, concerned with distances to the first nearest neighbor, NNA hinged on the behavior of the “nearest‐neighbour statistic,” R n , supposed to vary between 0 (complete clustering) and 2.15 (complete regularity). NNA and the wider body of work on point pattern analysis quickly gained in sophistication, meaning that this simple test statistic only featured briefly in the frontline research literature of geographical analysis. Nonetheless, given its easy‐to‐grasp logic, it became and remains a staple of quantitative geography textbooks for undergraduate students and statistical methods training in school‐level geography curricula. The purpose of this paper is to chart the history of NNA and its test statistic in academic geography, and to provide an analytical demonstration that the value of 2.15 has been mistakenly identified as an upper limit for the latter. Broader speculations are then offered about what may be learned from the history of how the discipline has handled this statistic.