Centre for Urban Network Evolutions, Aarhus University, Aarhus, Midtjylland Region, Denmark.
QuaDiHum Lab, Université libre de Bruxelles, Brussels, Brussels Capital Region, Belgium.
PLoS One. 2024 Sep 24;19(9):e0307743. doi: 10.1371/journal.pone.0307743. eCollection 2024.
Landscape research in archaeology has greatly benefited from the increasing application of computational methods over the last decades. Spatial statistical methods such as point pattern analysis have been particularly revolutionary. Archaeologists have used point pattern analysis to explore spatial arrangements and relations between 'points' (e.g., locations of artefacts or archaeological sites). However, the results obtained from these techniques can be greatly affected by the uncertainty coming from the fragmentary nature of archaeological data, their irregular distribution in the landscape, and the working methods used to study them. Furthermore, the quantification of uncertainty in spatial data coming from non-systematic surveys has never been fully addressed. To overcome this challenge, archaeologists have increasingly relied on applying advanced methods from statistics, data science, and geography. While the application of advanced methods from formal sciences will provide robustness to models based on uncertain datasets, as with uncertainty, robustness must be assessed in relation to the case study, the regional context, and the methods used. These issues are of great importance when the models from advanced methods are directly used to create narratives about past landscapes. In this paper, we gather previous research on uncertainty quantification in archaeology and formalize its best practices into a framework to assess robustness and uncertainty in spatial statistical models, particularly focusing on one commonly used in the discipline, i.e., the Pair Correlation Function. This framework allows us to understand better how incomplete data affect a model, quantify the model uncertainties, and assess the robustness of the results achieved with spatial point processes.
几十年来，随着计算方法在考古学领域应用的日益广泛，景观研究从中受益匪浅。空间统计方法，如点模式分析，尤其具有变革意义。考古学家已经使用点模式分析来探索“点”（例如，文物或考古遗址的位置）之间的空间排列和关系。然而，这些技术获得的结果可能会受到考古数据的不完整性、它们在景观中的不规则分布以及研究它们所采用的工作方法带来的不确定性的极大影响。此外，非系统调查得出的空间数据中的不确定性的量化从未得到充分解决。为了克服这一挑战，考古学家越来越依赖于应用统计学、数据科学和地理学中的先进方法。虽然来自正规科学的先进方法的应用将为基于不确定数据集的模型提供稳健性，但与不确定性一样，稳健性必须针对案例研究、区域背景和所使用的方法进行评估。当直接使用来自先进方法的模型来构建关于过去景观的叙述时，这些问题非常重要。在本文中，我们收集了考古学中不确定性量化的先前研究，并将其最佳实践形式化为一个框架，以评估空间统计模型的稳健性和不确定性，特别是重点关注该学科中常用的模型，即对关联函数。该框架使我们能够更好地了解不完整数据如何影响模型，量化模型不确定性，并评估空间点过程得出的结果的稳健性。
