Missing data in bioarchaeology I: A review of the literature

Abstract

Abstract Objectives Missing data are a frequent and unavoidable challenge in bioarchaeological research, yet researchers seldom make explicit statements about the bias and inferential limitations that missing data introduce into their studies. There are no guidelines for best practices for the treatment or reporting of missing data. As an initial step in taking stock and exploring approaches to missing data in bioarchaeology, this study reviews bioarchaeological publications to identify methods currently in use for addressing this significant problem. Materials and Methods Over 950 bioarchaeology articles (2011–2020) from four major anthropology journals were surveyed, searching for the terms “missing,” “absent,” “unobserv,” “replace,” and “imputat.” The 267 articles so identified were categorized into one of nine bioarchaeological subtopics and scored according to a set of six broad approaches for handling missing data. Results Results indicate that bioarchaeologists handle missing data in a variety of ways. Methods such as antimere substitution, listwise deletion and pairwise deletion are widely used. Subject subtopics favor different techniques for handling missing values. Bioarchaeological articles categorized as archaeology, pathology, and trauma used basic missing data approaches, while those such as biodistance and morphology more often employed advanced statistics. Despite the ubiquity of missing data, considerations of how they introduce bias were uncommon and standards for reporting were inconsistent. Conclusions These findings highlight areas in which bioarchaeologists can improve techniques for handling and reporting missing data. Greater attention to these shortcomings will increase the statistical rigor of the field.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 14 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.