Near presence cluster analysis: a new method for mobilizing presence / absence data for the analysis of spatial structure in archaeological survey

Abstract

Near Presence Cluster Analysis (NPCA) aids in the detection of spatial clustering among presence/absence data observed in irregularly distributed areal units. Archaeological surface surveys usually produce data that are aggregated into rectangular units. Depending on the sampling strategy, these units are often discontinuous and not distributed in a regular grid. Furthermore, depending on the collection strategy and the nature of artefacts found, much of the data generated is best seen as binary presence/absence data rather than continuous data. The nature of the data generated by archaeological surface survey can therefore make it difficult to apply established geostatistical methods when searching for spatial patterning. NPCA is designed to identify statistically significant clustering in precisely this type of difficult-to-analyse data. The heart of NPCA is the Near Presence (NP) score, a weighted average of the presences and absences, coded as 1s and 0s respectively, of a unit’s neighbours. A flexible approach to neighbour definition, either the n nearest units or all units within a certain radius, makes it applicable at any spatial scale and to any configuration of units. The weight of each neighbour is the inverse of the distance plus a constant, ensuring that nearer neighbours have more influence than farther ones. The significance of the NP score, whether it is high, low, or moderate, is determined for each unit individually through a permutation test. The R package ‘nearpresence’ performs all steps of NPCA and outputs spatial data with a variety of useful result attributes.