The estimation of biological sex is a primary source of information regarding unidentified skeletal individuals in bioarcheological and forensic contexts. This study aims to propose new metric standards for the estimation of sex using variables of the calcaneus and talus. An ancillary goal encompasses the creation of a web-based decision support system for the assessment of sex. Six measurements from the talus and nine from the calcaneus were collected from 180 adult individuals (93 females; 87 males) belonging to the Coimbra Identified Skeletal Collection. Logistic regression (LR), support vector machines (SVM), and a decision-tree algorithm were employed to develop models for sex prediction. Univariable sectioning points generated with a decision-tree algorithm yielded an accuracy under cross-validation from 78.3 to 82.2% with talar measurements, and from 73.6 to 86.4% with calcanei variables. Systematic error ranged from 0.2 to 34.1%. Univariable and multivariable models, produced with LR and SVM, correctly predicted sex in 85.0–91.3% of cases (bias from 0.3 to 4.3%). Obtained cross-validated accuracies obtained with the new models are similar to earlier results on the subject. The performance of multivariable model predictive is substantially superior, hinting the relevance of population-specific standards for sex estimation. The operationalization of these models in a free, user-friendly, web-application—CalcTalus (http://osteomics.com/CalcTalus/)—facilitates the probabilistic assessment of sex, providing performance metrics for the statistical templates.

生物性别的估计是有关生物考古学和法医鉴定中未鉴定的骨骼个体的主要信息来源。这项研究旨在提出一种新的度量标准，以使用跟骨和距骨的变量来估计性别。辅助目标包括创建用于评估性别的基于Web的决策支持系统。从属于Coimbra识别骨骼集合的180名成年个体（93名女性； 87名男性）中收集了距骨的六项测量值和跟骨的九项测量值。使用Logistic回归（LR），支持向量机（SVM）和决策树算法来开发性别预测模型。通过决策树算法生成的单变量剖分点在进行交叉验证时，使用距骨测量，其准确度从78.3％提高到82.2％，加上calcanei变数则介于73.6％至86.4％之间。系统误差范围为0.2到34.1％。由LR和SVM生成的单变量和多变量模型可以正确预测85.0–91.3％的病例中的性别（偏倚从0.3％到4.3％）。使用新模型获得的交叉验证精度与该主题上的早期结果相似。多变量模型预测的性能明显优越，这暗示了针对性别的特定人群标准的相关性。这些模型可在免费的，用户友好的Web应用程序CalcTalus（http://osteomics.com/CalcTalus/）中进行操作，从而有助于对性别进行概率评估，从而为统计模板提供性能指标。LR和SVM产生的性腺，在85.0–91.3％的病例中正确预测了性别（偏倚从0.3％到4.3％）。使用新模型获得的交叉验证精度与该主题上的早期结果相似。多变量模型预测的性能明显优越，这暗示了针对性别的特定人群标准的相关性。这些模型可在免费的，用户友好的Web应用程序CalcTalus（http://osteomics.com/CalcTalus/）中进行操作，从而有助于对性别进行概率评估，从而为统计模板提供性能指标。LR和SVM产生的性腺，在85.0–91.3％的病例中正确预测了性别（偏倚从0.3％到4.3％）。使用新模型获得的交叉验证精度与该主题上的早期结果相似。多变量模型预测的性能明显优越，这暗示了针对性别的特定人群标准的相关性。这些模型可在免费的，用户友好的Web应用程序CalcTalus（http://osteomics.com/CalcTalus/）中进行操作，从而有助于对性别进行概率评估，从而为统计模板提供性能指标。多变量模型预测的性能明显优越，这暗示了针对性别的特定人群标准的相关性。这些模型可在免费的，用户友好的Web应用程序CalcTalus（http://osteomics.com/CalcTalus/）中进行操作，从而有助于对性别进行概率评估，从而为统计模板提供性能指标。多变量模型预测的性能明显优越，这暗示了针对性别的特定人群标准的相关性。这些模型可在免费的，用户友好的Web应用程序CalcTalus（http://osteomics.com/CalcTalus/）中进行操作，从而有助于对性别进行概率评估，从而为统计模板提供性能指标。