[Paper Review] WARN: an R package for quantitative reconstruction of weaning ages in archaeological populations using bone collagen nitrogen isotope ratios

Takumi Tsutaya, Minoru Yoneda|arXiv (Cornell University)|Apr 9, 2013

Forensic Anthropology and Bioarchaeology Studies56 references3 citations

TL;DR

This paper introduces WARN, an open-source R package that enables quantitative, probabilistic reconstruction of weaning ages in archaeological subadults using bone collagen nitrogen isotope ratios (δ¹⁵N). By integrating newly estimated subadult bone collagen turnover rates and an approximate Bayesian computation (ABC) framework, WARN provides posterior probabilities and credible intervals for key parameters like weaning start/end ages, maternal-to-infant δ¹⁵N enrichment, and weaning food δ¹⁵N values, overcoming the subjectivity of prior visual assessments.

ABSTRACT

Nitrogen isotope analysis of bone collagen has been used to reconstruct the breastfeeding practices of archaeological human populations. However, weaning ages have been estimated subjectively because of a lack of both information on subadult bone collagen turnover rates and appropriate analytical models. Here, we present a model for analyzing cross-sectional delta-15N data of subadult bone collagen, which incorporates newly estimated bone collagen turnover rates and a framework of approximate Bayesian computation. Temporal changes in human subadult bone collagen turnover rates were estimated anew from data on tissue-level bone metabolism reported in previous studies. A model for reconstructing precise weaning ages was then developed and incorporating the estimated turnover rates. The model is presented as a new open source R package, WARN (Weaning Age Reconstruction with Nitrogen isotope analysis), which computes the age at the start and end of weaning, 15N-enrichment through maternal to infant tissue, and delta-15N value of collagen synthesized entirely from weaning foods with their posterior probabilities. A precise reconstruction of past breastfeeding and weaning practices over a wide range of time periods and geographic regions could make it possible to understand this unique feature of human life history and cultural diversity in infant feeding practices.

Motivation & Objective

• To address the lack of objective, quantitative methods for estimating weaning ages from archaeological subadult bone collagen δ¹⁵N data.
• To overcome the limitations of previous models that ignored subadult bone collagen turnover rates and treated key parameters as fixed.
• To develop a statistical framework that incorporates uncertainty by estimating posterior probabilities and credible intervals for weaning parameters.
• To provide a reproducible, open-source tool (WARN) for researchers to analyze cross-sectional δ¹⁵N data from archaeological populations.

Proposed method

• Re-estimated temporal changes in subadult bone collagen turnover rates using data from tissue-level bone metabolism studies.
• Formulated a dynamic model of δ¹⁵N change in subadult bone collagen that accounts for age-dependent turnover rates.
• Incorporated variable parameters: maternal-to-infant δ¹⁵N enrichment and δ¹⁵N of weaning foods as unknowns to be estimated.
• Applied approximate Bayesian computation (ABC) to compute posterior distributions and credible intervals for weaning start/end ages, enrichment, and weaning food δ¹⁵N.
• Implemented the model as a user-friendly R package (WARN) with visualization tools for results interpretation.
• Validated the model using the Spitalfields archaeological population dataset, demonstrating consistent posterior probability distributions for key parameters.

Experimental results

Research questions

• RQ1 What is the range of weaning start and end ages in the Spitalfields population, and what is the probability of these estimates?
• RQ2 How much δ¹⁵N enrichment occurs from maternal to infant tissues, and what is the credible interval for this enrichment factor?
• RQ3 What is the estimated δ¹⁵N value of collagen synthesized entirely from weaning foods in the studied population?
• RQ4 How do age-dependent bone collagen turnover rates influence the accuracy of weaning age estimation?
• RQ5 To what extent does the model improve upon previous subjective or point-estimate-based methods in reconstructing weaning practices?

Key findings

• The model successfully estimated a 95.6% joint posterior probability that weaning began between 0.0 and 1.2 years and ended between 1.2 and 2.0 years of age in the Spitalfields population.
• The marginal posterior probability for δ¹⁵N enrichment from maternal to infant tissues falling within the 1.6–2.4‰ range was 0.967.
• The marginal posterior probability for the δ¹⁵N value of collagen from weaning foods being within the 12.4–13.0‰ range was 0.975.
• The model's incorporation of age-specific bone collagen turnover rates significantly improves the accuracy of weaning age estimation compared to models assuming constant turnover.
• The package provides full uncertainty quantification through posterior probability distributions, enabling researchers to assess the reliability of reconstructed weaning parameters.
• The WARN R package enables reproducible, transparent, and statistically robust analysis of subadult δ¹⁵N data across diverse archaeological and paleoanthropological contexts.