Sex assessment using the radius bone in a French sample when applying various statistical models

Area of Science:

• Forensic Anthropology
• Bioanthropology
• Osteology

Background:

• Sex estimation from skeletal remains is crucial in forensic anthropology.
• The radius bone offers a viable alternative when pelvis or femur are unavailable.
• Previous research has explored various methods for sex estimation from skeletal elements.

Purpose of the Study:

• To assess sexual dimorphism in radius bone measurements within a French population.
• To compare the accuracy of classical statistical methods versus machine learning algorithms for sex estimation.
• To identify key radius measurements predictive of sex.

Main Methods:

• Utilized 78 left radii (36 male, 42 female) from a French sample.
• Performed sixteen distinct bone measurements.
• Applied various modeling techniques: Linear Discriminant Analysis (LDA), Flexible Discriminant Analysis (FDA), Regularised Discriminant Analysis (RDA), Penalised Logistic Regression (PLR), Random Forests (RF), and Support Vector Machines (SVM).

Main Results:

• Initial models achieved over 94% classification accuracy.
• Variable selection enhanced accuracy to 97%.
• Key predictive measurements included proximal radius (head/neck diameters), distal radius (distal epiphysis width), and maximum bone length.

Conclusions:

• The radius bone is a valid alternative for skeletal sex estimation, comparable to pelvis and femur.
• Non-classical statistical models (machine learning) offer a novel and accurate approach.
• Results are based on an aged sample, requiring cautious extrapolation.