Comparison of Visual and Statistical Pair Matching of the Humerus in a Large Assemblage

Abstract

Determining whether antimeres are consistent with each other is important in estimating the minimum number of individuals in an assemblage, and this is often done with a simple visual comparison. When there are a small number of individuals, this task is generally straightforward and quick. However, as the number of individuals increases, the ability to make visual comparisons of all elements becomes cumbersome. In these cases, osteometric sorting has become a useful tool for creating shortlists of possible antimeres. Shortlists can save time by providing a smaller list of antimeres to visually compare. However, the risk of the true pair being rejected or not being reflected on the shortlist is a concern. Therefore, this study seeks to determine which statistical methods reject true pairs the least often and how these methods compare to visual pair matching. The default statistical parameters set by OsteoSort that utilize the mean of the summed left–right differences (D) performed better than using the mean of the summed absolute left–right differences (| D |). Utilizing specific measurement combinations in osteometric sorting tests, rather than all possible measurements, was only partially more successful. The true pair was excluded or rejected in statistical tests more often than true pairs being missed in a visual pair matching (VPM) study. Therefore, it is recommended to conduct VPM if the time, space, and resources allow.