Scholarlink

<p>The ability to compare finds assemblages quantitatively underlies many problems in archaeology, such as whether surface or subsurface assemblages reflect one another in landscape archaeology, or whether two or more sets of contexts saw the same distributional patterns of artifact use and deposition. This paper advances a statistically informed approach that relies on the Cressie-Read power-divergence statistic, which generalizes Pearson’s chi-squared and <i>G</i> tests. This approach also advocates for a focus on effect size. By basing the evaluation of what are “homogenous” distributions of artifacts on a partition of pairs of contexts within a project, measures of the similarity of assemblages are attained for both count and presence/absence data, whose robustness can be assessed with a leave-one-out resampling routine. Rather than using arbitrary benchmarks for what constitutes “large” or “small” effect sizes, the magnitude of similarity for related assemblages is expressed directly via comparison with unrelated assemblages.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluating the Relationship Between Surface, Subsurface, and Stratigraphic Assemblages

  • 期刊论文

  • 发表日期：

    2026年8月4日

查看全文
  • Stephen A. Collins-Elliott

摘要

The ability to compare finds assemblages quantitatively underlies many problems in archaeology, such as whether surface or subsurface assemblages reflect one another in landscape archaeology, or whether two or more sets of contexts saw the same distributional patterns of artifact use and deposition. This paper advances a statistically informed approach that relies on the Cressie-Read power-divergence statistic, which generalizes Pearson’s chi-squared and G tests. This approach also advocates for a focus on effect size. By basing the evaluation of what are “homogenous” distributions of artifacts on a partition of pairs of contexts within a project, measures of the similarity of assemblages are attained for both count and presence/absence data, whose robustness can be assessed with a leave-one-out resampling routine. Rather than using arbitrary benchmarks for what constitutes “large” or “small” effect sizes, the magnitude of similarity for related assemblages is expressed directly via comparison with unrelated assemblages.